Introduction

Dementia encompasses deteriorations in several cognitive domains [8, 36, 103], see [40]; it has been re-defined as the differential manifestation of deteriorating brain functions over time as a part of aging due to cell deaths in the brain caused by neurodegeneration or any other disease [117]. However, because of recent research data, dementia is not only caused by ‘neuronal cell death’/cell loss [91, 104, 151], but predominantly by dysfunction and loss of synapses [43, 60, 128, 129, 140, 141] that have been demonstrated in early Alzheimer disease (AD) [123] and in dementia with Lewy bodies (DLB) [83], and by cholinergic neuronal and axonal abnormalities that also are present in aging and AD [48]. These changes cause disconnections of important nervous circuitries [14, 35, 68, 114] that have been demonstrated in vivo in early AD [152].

The aim of this review is to discuss the diagnostic validity of currently used neuropathological criteria particularly of neurodegenerative dementias and their limitations as well as to give recommendations for future clinicopathological research.

Research and consensus criteria for the clinical diagnosis of the major dementing disorders exist and have recently been revised, e.g., the NINCDS-ADRDA and DSM-IV-TR criteria for AD [36, 37], Parkinson disease dementia (PDD) [42], DLB [46, 97], frontotemporal lobe degeneration (FTLD) [77, 90, 109], vascular dementia or vascular cognitive disorder (SCADDTD [26]; NINDS-AIREN [106, 125, 126, 155]), mixed dementias [71], and others. Although the NINCDS-ADRDA criteria [98] combined with neuropsychologic assessment are still valid for 88–90% of AD [80], they do not exclude additional Lewy body and other pathologies [122]. Combination of clinical data with the fusion of different biomarkers has already improved the clinical diagnostic accuracy of AD up to 96%, while their sensitivity and specificity versus other dementias is still lower. A combination of the best CSF and MRI data will lead to a more precise diagnostic prediction [30].

Advantages and pitfalls of neuropathological diagnostic criteria

At present, several diagnostic guidelines for the neuropathological diagnosis of most dementing disorders are used, relying on the qualitative, semiquantitative, and topographic assessment of morphological and bio/histochemical sign posts, in particular specific protein inclusions in neurons, glia and other cells. The neuropathology of dementia has been suggested to appear as a “science of inclusion bodies” [39], see also [40]. A classification of degenerative dementias based on currently known biochemical markers is given in Table 1.

Table 1 Biochemical classification of degenerative dementias (modified after [39, 68])
Full size table

For α-synucleinopathies, in particular Lewy body disorders, in addition to assessment criteria [5], staging/classification systems based on the semiquantitative assessment of the distribution and progression pattern of α-synuclein (αSyn) pathology are used that are suggested to indicate a predictable sequence of lesions [22, 24]. They allow a distinction of three major phenotypes of LB disease—brainstem predominant, transitional/limbic and diffuse cortical [89, 97], while AD with amygdala Lewy bodies is considered a distinct form of α-synucleinopathy [147]. Recent clinicopathologic studies, although partly confirming these systems, have shown that between 6.3 and 47% of cases did not follow the proposed staging pattern of αSyn pathology [75, 76, 116]. On the other hand, 30–55% of elderly subjects with widespread/cortical αSyn pathology revealed no definite neuropsychiatric symptoms or were not classifiable [1, 17, 71, 72, 86, 159], and the criteria for categorization of Lewy-related pathology in patients with dementia had to be modified [1, 46, 86]. PDD and DLB, sharing many clinical and morphological features and believed to form a continuum within the spectrum of LB diseases [11, 55, 97], have been shown to differ by more severe diffuse amyloid load in the striatum of DLB [69, 87], recently confirmed by in vivo studies using 11CPIB-PET [41, 54], as well as by more frequent LB affection of the hippocampal CA 2–3 subareas [76]. The predictive value of striatal Aβ pathology, with regard to cognitive impairment is still controversial [1].

For FTLD, nowadays suggested to be the third or forth most frequent cause of dementias, criteria for neuropathological diagnosis are based on the basic biochemical markers [25, 77, 90, 105] including the novel TDP-43 neurodegenerative proteinopathies [45]. Recently, a small novel group of FTLD patients with clinical features that overlap with DLB has been identified, which morphologically were consistent with TDP-43 proteinopathy (FTLD with ubiquitin-only lesions, type I) [29, 149].

Despite various proposals for a categorization of major cerebrovascular lesions [78], a harmonization of the criteria and techniques for the assessment of cerebral lesions of presumable/possible vascular origin in cognitively impaired is necessary [115]. Because of its high variability of morphological findings and multifactorial pathogenesis of vascular cognitive impairment, no generally accepted morphologic scheme for quantitating cerebrovascular lesions and no validated neuropathological criteria for vascular dementia have been established to date (see [70]). The same holds for so-called “mixed dementia” [71, 73]. Recently, BrainNet Europe II (BNE) has constituted a “vasculopathy” reference group to harmonize the assessment of vascular pathology similar to previous interlaboratory studies for the morphological assessment of AD-related and αSyn pathologies [5, 6, 23].

In contrast to most dementing conditions that typically develop over years, rapidly progressive dementia being quickly fatal, is one of the most challenging neuropathological problems. The differential diagnosis is often widely ranging, and in addition to frequent sporadic or genetic prion diseases includes rapidly progressing neurodegenerative tauopathies and synucleinopathies, autoimmune conditions infectious, toxic-metabolic and neoplastic diseases [47]. In these and other unclear conditions, brain biopsy may play a role, although frequent biopsy findings in dementia are non specific [153].

Specific problems in the diagnosis of AD

Histopathological examination of the brain establishes that AD-related lesions are present in sufficient densities to distinguish AD from age-related lesions and allows detection of other dementing disorders [39]. Although the interlaboratory comparison of neuropathological assessment of AD when using standardized criteria showed reasonable interrater agreement [3, 4, 6, 13, 38, 56, 95, 102, 108, 109], no one set of histopathological criteria for AD has been uniformly accepted by neuropathologists [156]. These algorithms that only consider the classical “plaque and tangle” phenotype of AD do not recognize other dementias and AD subtypes, e.g., the “plaque predominant” type with abundant amyloid plaques, no or very little neuritic pathology restricted to the hippocampus and abnormal phosphorylated tau in neocortical pyramidal cells but lacking overt tangle formation, accounting for 3.5–8% of demented subjects over age 85 years [66, 139, 144, 145], the “tangle dominant type” with 3+4R tau pathology often restricted to the limbic system, absence of neuritic plaques and no or very little amyloidosis, accounting for 5–7% of oldest-olds [73], and the Lewy body variant of AD, with cortical and subcortical Lewy bodies, often associated with severe AD pathology [57]. Recent studies showed different cytoskeletal alterations in familial AD (FAD) due to PSEN1 mutations and sporadic AD [157]. Furthermore, since standard neuropathological metrics for tangles and neuritic plaques [21, 62, 102] are usually semiquantitative and, according to the BNE consortium, good agreement can be reached in the neuropathological diagnosis only when the lesions are substantial, e.g., when they have reached isocortical structures (Braak stage V–VI with absolute agreement 91%). In contrast, for mild lesions the agreement was poorer (Braak stage I–II, agreement 50%) [6], thereby limiting the ability to make accurate correlation of antemortem cognitive status and the severity of morphological findings. Although the sensitivity and specificity of the NIA-RI criteria is suggested to be 90%, only 40–57% of the brains of patients with the clinical diagnosis of probable AD show “pure” AD pathology (Table 4). Thus, their predictive value may be reduced to 38–44% [20].

Table 4 Mixed pathologies frequency in demented elderly
Full size table

Diffuse and neuritic plaques and some amounts of tau-positive neuritic pathology in the limbic system relatively frequently occur in cognitively normal elderly [10, 15, 18, 32, 34, 61, 68, 81, 120, 130]. Although cognitively unimpaired subjects may show some or even considerable AD-related pathology [94, 137, 138], in general, the number of isocortical tangles correlates best with clinical dementia severity [12, 15, 19, 92, 110, 131, 144, 145], and patterns of gray matter loss associated with tangle pathology is an appropriate in vivo surrogate indicator of AD pathology [154]. The predictive value of widespread tau pathology (Braak stages V–VI) for dementia is high [1], while others found that both diffuse and neuritic plaques, rather than tangles in neocortical regions distinguish non-demented and AD subjects with high sensitivity and specificity [96].

Neuropathology of AD in dementia in the oldest-old differs considerably in both intensity and distribution from that in younger age groups [49, 50, 150]. Significantly increased densities of neuritic plaques and NFTs are absent in demented patients over age 85–90 years [52, 59, 64, 110, 119, 121, 136], and there is a considerable overlap in the pathologies found in the demented and non-demented [158]. A high percentage of demented aged persons aged 80+ do not meet the pathological criteria of AD or were classified as “dementia of unknown etiology” [33, 65].

Another important point is the frequent presence of confounding processes in the aged brain that coexist with AD, e.g., cerebrovascular disease, Lewy body pathology, argyrophilic grain disease, hippocampal sclerosis etc, about two-thirds of aged human brains containing non-AD-type neuropathology [107, 110]. The frequency of mixed pathologies in demented elderly is shown in Table 4, and was recently confirmed in over 3,300 autopsy cases of aged demented individuals [82]. Since 50–85% of the brains of persons who die aged 80–90+ old show appreciable cerebrovascular lesions [118], a specific problem is the impact of cerebrovascular disease in relation to AD pathology [16, 27, 28, 51, 110, 134, 146]. The burden of vascular and AD type lesions are considered to be independent of each other, and are consistent with an additive or synergistic effect of both types of lesions on cognitive impairment [67, 71, 72, 84, 132]. The thresholds for vascular and degenerative lesions in distinguishing “pure” VaD or AD from mixed cases have been critically discussed recently [53, 160]. AD pathology alone more frequently accounts for dementia than both microscopic and macroscopic infarcts [146], and in advanced or full-blown stages of AD concomitant small vascular lesions do not significantly influence the overall state and progression of cognitive decline, the severity and extent of AD pathology overwhelming the effects of cerebrovascular disease [27, 71–73, 85]. Nevertheless, it should be borne in mind that all additional pathologies may interact. Therefore, the reliability and clinical relevance of the current diagnostic criteria need better qualification and validation.

Conclusions

Adequate use of morphological markers (tau, amyloidosis, synuclein, synaptic and other proteins) as well as biochemical markers usually identify the vast majority of cases with dementia. However, due to variable overlap, these changes may fail to distinguish between cognitively intact aged subjects from those with mild cognitive impairment or preclinical or mild AD. In particular, these latter groups show a wide variety in the intensity and pattern of AD-related lesions. Although they often differ from “normal” aging, only a small proportion of cognitively intact aged are free of AD pathology, while up to 50% may show AD-related changes or even definite AD pathology [10, 15, 18, 32, 34, 61, 68, 81, 120, 130]. Additional difficulties arise from frequent coexistence with other pathologies that may have an additive or synergistic effect. Similar difficulties arise in other neurodegenerative dementias, in particular those with genetic background. The question, whether the neuropathological “gold standard” is dead has been discussed in respect to genetical neurodegenerative disorders that offer a rare opportunity to test the validity of neuropathological diagnostic criteria. Implications regarding an autosomal dominant disorder (PARK 8) in which four different morphological diagnoses were found at autopsy were discussed. It was suggested that just as there is currently no clinical “gold standard” for Parkinson disease, in certain circumstances there is also no pathological “gold standard”, and only the combination with genetic studies may provide definitive arbitration of validity of clinical and pathologic diagnostic criteria [148, 149]. In view of these difficulties, the present author feels not competent enough to present or propose “new” criteria for the neuropathological diagnosis of neurodegenerative and dementing disorders and, thus, to show a “simple” way out of the current “chaos” regarding histological diagnosis of dementia and their clinical implications. Further studies using methodological and validated protocols and harmonized techniques are required to increase the accuracy and reproducibility of neuropathological diagnosis. Keeping all the pros and cons in mind, one has to conclude that, if robust correlations between clinical course and morphological changes will be confirmed by future correlative studies, the neuropathological classification and staging systems will need to be revised accordingly. There will be a long and difficult way out of the swamp.