Of all the neuropathologic features that we evaluated, the odds of a clinical diagnosis of AD were most strongly associated with neurofibrillary tangles across all age groups; however, the strength of the association between clinical AD diagnosis and neurofibrillary tangles was significantly weaker among the oldest old, even when the presence of multiple pathologies was considered. Neuropathologic features, whether AD pathology only or the aggregate of all coexisting pathologic features, differentiated people with clinical AD from those with normal cognition more poorly with increasing age; however, consideration of non-AD pathology improved the precision of differentiation in all age groups.
It has long been understood that clinical cognitive diagnoses do not always correspond with neuropathologic findings postmortem.21
That is, some people with normal cognition at death have neuropathologic changes consistent with a pathologic diagnosis of AD or other dementia.21,22
Conversely, some people who are diagnosed with dementia show no obvious neuropathologic cause upon autopsy.23
However, this discord between clinical diagnosis of AD and neuropathologic features, and particularly AD pathology, seems to widen at higher ages.7
In line with most previous studies,6,7,13
the strength of the association between clinical diagnosis of AD and neurofibrillary tangles was weaker among the oldest old than the young old, even when the presence of coexisting neuropathologic features was taken into account. The reason why clinical AD diagnosis was more weakly associated with neurofibrillary tangles with greater age is unclear; however, there are several possible contributing factors. First, the level of cognitive impairment among those with AD was less severe among the oldest old than among the young old in our sample, as evidenced by higher MMSE scores prior to death (). For that reason, the weaker relationship between neurofibrillary tangles and clinical cognitive diagnosis among the oldest old may be an artifact of disease severity, rather than age group. However, this is unlikely given that Haroutunian et al.6
found that the association between AD pathology and dementia severity was also diminished among the oldest old. Second, the frequency of neurofibrillary tangles among those with normal cognition was higher in the oldest old than in other age groups. This suggests that some of the oldest old who were classified with normal cognition may have subclinical AD. Third, in secondary analyses, the oldest old were more likely to have neurofibrillary tangles without concomitant neuritic plaques—in our sample 25% of the oldest old had high levels of neurofibrillary tangles but low levels of neuritic plaques compared to 10% of those 70 to 74 years (p
< 0.001). This corroborates previous observations that the oldest old are more likely to have a condition coined as senile dementia with tangles, which is characterized by massive deposition of neurofibrillary tangles in limbic areas without concomitant neuritic plaques.24
It is possible that neurofibrillary tangles may be less likely to cause sufficient cognitive impairment to be diagnosed with clinical AD without the additional contribution of neuritic plaques. Finally, coexisting microvascular disease, which contributes to cognitive impairment, may have been more severe among older age groups, where severity may be important for predicting clinical outcomes.25
In the NACC database, microvascular disease was only reported as present or absent so the reports were insufficient to capture an effect of severity.
Unlike the relationship with neurofibrillary tangles, clinical AD was similarly associated with other neuropathologic features across age groups when we controlled for coexisting neuropathologic features. Even though the prevalence of vascular pathologies increased with age, the odds of AD based on vascular pathologies did not because people with normal cognition also had an increase in the frequency of vascular pathologies with age, though this was not statistically significant. Furthermore, a trend for diminished association between clinical AD and neuritic plaques and Lewy bodies, which was observed in other studies,6,7,13,26
was attenuated when coexisting neuropathologic features were considered. This suggests that the preliminary observation of a trend toward weaker association between clinical AD and Lewy bodies and neuritic plaques in older age groups in this and other studies may be partially due to differing levels of coexisting neuropathologic features that also contribute to cognitive impairment across age groups (such as neurofibrillary tangles and microvascular disease).8,27
In this study, the average number of coexisting neuropathologic features did not differ by age, in contrast to previous studies,28
so it is likely the combinations of neuropathologic features that were of primary importance.
In this study, neuropathologic features differentiated people with clinical AD from those with normal cognition less accurately among the oldest old relative to the young old in ROC analyses, though non-AD pathology significantly improved the differentiation accuracy across all age groups. Why the precision by which neuropathologic features differentiated those with from those without a clinical diagnosis of AD among the oldest old was worse than among the young old is unclear. It is possible that the clinical diagnosis is less accurate in this group because neuropsychological assessments often do not have norms that are specific to the oldest old. Alternatively it may be due to survival bias, where people who survive to very old ages are also more likely to employ successful neurocompensation techniques, whether genetically predetermined or behaviorally mediated, to maintain cognition despite significant pathologic changes. It is also possible that there are additional neuropathologic features that are important to consider among the oldest old that are not accounted for here. For example, cerebral atrophy and synaptic protein loss, not recorded in the NACC database, are understood to be associated with cognition in the oldest old and the young-old.13,29
In particular, the frequency of cerebral atrophy increases with age.28
Importantly, though worse among the oldest old, the neuropathologic features differentiated those with from those without a clinical diagnosis of AD very well in all age groups. Even among the oldest old, neuropathologic features could determine the likelihood of a clinical cognitive diagnosis to 81% accuracy. That is, in any pairing, there was 81% chance that an individual with a clinical diagnosis of AD had greater neuropathologic load than someone with normal cognition.
Our study has several strengths. We had a large sample with standardized neuropathologic reports. In addition, all participants had clinical cognitive diagnosis determined at an ADC. However, our study also has some limitations. Although all ADCs reported common neuropathologic measures, the methods to determine each measure may have differed between sites and by year. In addition, the rating of vascular pathology was imprecise, a common problem faced in neuropathologic examinations,30
and was marked only as present or absent. We also did not have a measure of cerebral atrophy. In addition, although we had clinical cognitive diagnosis, we did not have a continuous measure of cognition close to death for the entire sample. Finally, those included in the NACC database are unlikely to represent the general AD population31
; the database includes fewer people with severe AD than would be found in the general AD population. In addition, participants are primary white and highly educated.
The oldest old accounted for approximately 40% of dementia cases by the mid-1990s. Since the oldest old are the fastest growing segment of the population, this proportion is expected to rise in the coming decades. The relationship between clinical diagnosis of AD and neurofibrillary tangles is attenuated among the oldest old and neuropathologic features do not differentiate people with from people without a clinical AD diagnosis as well in this age group compared to the young old. The explanation is not clear but it is possible that additional neuropathologic features need to be taken into account or that these oldest old survivors are better able to cope with neuropathology to maintain cognition despite pathologic changes in the brain. Future studies should further investigate the impact of additional coexisting pathology on cognition in the oldest old.