The literature search was conducted on the PubMed, ScienceDirect, Scopus, and Web of Science databases. The search was limited to articles published between January 2011 and December 2021. The search strategy was as follows: [Alzheimer’s disease] AND [atypical Alzheimer’s disease] OR [Alzheimer variants] OR [amnestic Alzheimer disease] OR [logopenic variant primary progressive aphasia] OR [posterior cortical atrophy] OR [frontal variant Alzheimer disease] OR [corticobasal degeneration] OR [corticobasal syndrome] OR [neuropsychological assessment] OR [cognition] OR [cognitive function] OR [language impairment] OR [memory] OR [executive function] OR [biomarkers] OR [neuroimaging biomarkers]. All search terms were adapted to each database. The complete search strategy for each database is included in the supplementary material (Appendix 2).

We applied the following inclusion criteria for screening by title and abstract: (a) analytical observational studies of diagnostic tests, reporting the neuropsychological assessment and biomarker test results; (b) studies written only in English; and (c) including patients aged at least 60 years with a clinical diagnosis of AD-typ, lvPPA, PCA, bvAD, and/or CBS. The exclusion criteria applied were the following: (a) editorials, experimental studies, systematic reviews (with or without meta-analyses), protocols, and theses; (b) including patients with FTD, Lewy body dementia, vascular dementia, or mixed dementia; and (c) including patients with history of psychiatric disorders (e.g., depression).

Studies were imported to the Mendeley software (version 1.19.4) to eliminate duplicate articles. Subsequently, 3 reviewers (CFR, FGB, and CRB) applied the inclusion and exclusion criteria to the titles and abstracts of all articles. In the event that decisions could not be made based solely on the title and abstract, the reviewers recovered the full texts of the articles. A fourth researcher (DTR) participated in selection consensus.

To evaluate the methodological quality and risk of bias of the 31 articles selected, we used the QUADAS-2,14 an instrument designed and validated for the independent evaluation of methodological quality and risk of bias in studies of diagnostic accuracy. This instrument evaluates 4 key domains: patient selection, the index test, the reference standard, and the patient flow and timing of the study (the latter 2 aspects are grouped together into the domain “flow and timing”); given the type of studies included, the item “reference standard” was not applicable. Studies were categorised into one of 3 risk categories: high, low, or unclear.

Tables 1 and 2 present a narrative synthesis of our findings. Table 1 summarises the general characteristics of the studies, including year of publication, study population, mean age, and the instruments/measures used. Table 2 presents the neuropsychological and neural profiles of the typical and atypical forms of AD.

The article selection process is illustrated in the PRISMA flow chart (Fig. 1).13 We identified the titles and abstracts of 5389 articles, 1227 of which were duplicates. After applying the inclusion/exclusion criteria to the 2162 articles screened, 2077 articles were eliminated: 1246 due to the type of study, 813 due to the type of population, and 18 due to the language of publication. Thus, the full texts of 85 articles were analysed to assess their eligibility. Of these, 54 articles were eliminated: 12 due to the type of population and 42 due to the type of evaluation. Finally, a total of 31 articles were included for review.

Fig. 1.

PRISMA flow diagram.

*Consider, if feasible to do so, reporting the number of records identified from each database or register searched (rather than the total number across all databases/registers).

**If automation tools were used, indicate how many records were excluded by a human and how many were excluded by automation tools.

(0.48MB).

Table 1 shows the general characteristics of the 31 articles included for review.15–45 The years in which the highest number of articles were published were 2019 (5 articles), 2020 (5 articles), and 2021 (4 articles); in contrast, only 2 articles per year were published between 2011 and 2018, with the exception of 2014, in which 1 article was published. The total sample included 2133 elderly patients, with a mean age of 66.2 years. The most frequently used neuropsychological assessment tools and measures for the different profiles of AD included the Addenbrooke’s Cognitive Examination (ACE), Boston Naming Test (BNT), Clinical Dementia Rating Scale (CDR), F-A-S verbal fluency test (FAS), Mini–Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Rey-Osterrieth Complex Figure (ROCF), Trail-Making Test (TMT), Visual Object and Space Perception (VOSP) battery, and the Wechsler Adult Intelligence Scale (WAIS). The most commonly used imaging techniques were MRI, PET, and SPECT.

Table 2 shows how the neuropsychological profiles of the different subtypes of AD share deficits in episodic memory, working memory, executive function, language (verbal fluency), and visuospatial/visuoconstructive skills. In turn, the neural correlates identified share patterns of atrophy in the frontal, temporoparietal, and occipital areas, as well as the cuneus and precuneus.

The 31 articles selected were analysed according to the QUADAS-2 guidelines (Fig. 2).14 In the bias analysis, all articles were categorised as low risk in the flow and timing domain; in the index test domain, 94% were classed as low risk and 6% as unclear risk; and in the patient selection domain, 71% were categorised as unclear risk and 20% as low risk. In turn, in the analysis of methodological quality, 58% presented high quality and 42% unclear quality for the applicability of results in the index test domain, while 77% of studies presented unclear quality and 23% high quality for applicability of results in the patient selection domain.

Fig. 2.

Risk of bias and methodological quality in the studies reviewed.

(0.14MB).

Several articles46–49 report similar findings regarding memory deficits in AD-typ, highlighting early difficulties with the registration and consolidation of verbal and visual information. Meng et al.48 propose using the Common Objects Memory Test (COMT)50 for the assessment of episodic memory, due to its sensitivity in early stages. The semantic and episodic memory deficits reported in AD-typ are consistent with those described in previous studies.51,52

Executive function deficits are common in the early stages of AD-typ, and have been proposed as a predictor of development and progression.53 Kashiwa et al.54 describe how executive dysfunction promotes significant deficits in the inhibitory response. One study emphasised that this dysfunction can also increase symptoms of apathy and even the appearance of affective disorders, complicating the initial diagnosis of AD-typ and bvAD.55

Semantic fluency is severely affected in AD-typ, unlike phonological fluency.56,57 Henry et al.51 note that phonological fluency does not constitute an additional characteristic in this disorder; however, other authors58–60 suggest that it should be considered a predictor in differential diagnosis. Recent studies61–63 suggest evaluating spontaneous language in early AD-typ, with specific emphasis on prolonged pauses during speech.

Similarly to our own findings, Eikelboom et al.64 report that lvPPA presents with deficits in episodic and working memory. Win et al.65 suggest that the significant decline in verbal and autobiographical episodic memory is similar to that reported in AD-typ. Other studies66,67 suggest that the decline in short-term and working memory is minimal in early stages.

Studies suggest that executive function is frequently affected not only in lvPPA, but also in the semantic and non-fluent/agrammatic variants of PPA.68,69 Executive dysfunction in lvPPA affects information processing speed, sustained attention, and visuospatial/visuoconstructive skills.25,70–72 Similarly, Kamath et al.73 report that the magnitude of the deficit may be equally significant as that of linguistic alterations in lvPPA.

The most relevant deficits in lvPPA are problems with phonological fluency, phonological paraphasia, difficulties with sentence repetition and comprehension, and word-finding difficulties.74–76 Although our results are consistent with these difficulties, the literature also suggests an alteration in semantic fluency, similar to that reported in AD-typ.77,78

Deficits are reported in the coding and retrieval of verbal information in early stages of PCA,79–81 similar to those reported in AD-typ.82 Memory deficits are not limited to working memory and conceptual reasoning,83–85 but rather also affect episodic and autobiographical memory, which constitutes a challenge for differential diagnosis with AD-typ.86,87 Kas et al.36 showed that episodic memory performance in PCA is better in the first 3 years after diagnosis.

Putcha et al.40 report that the executive deficit in PCA has an impact on verbal episodic memory, and Li et al.88 note that it is more severe in PCA than in AD-typ, also reporting significant involvement of language and visuospatial, visuoconstructive, and visuoperceptual skills. Several other studies83,89–91 describe deficits in visuospatial skills in PCA, which in turn lead to such progressive disorders as visual agnosia,92 oculomotor alexia,93 acalculia, and simultanagnosia.94

Anomia is frequent and marked in PCA, manifesting early in the course of the disease.94,95 Migliaccio et al.96 found no significant differences in naming and phonological fluency performance between PCA and lvPPA. Similarly, Crutch et al.97 report alterations in both groups in auditory processing, repetition, and phonological fluency, with less severe involvement of spontaneous language and comprehension in PCA.

Lehingue et al.98 studied patients with bvAD and AD-typ, reporting poorer episodic memory performance in the latter group. Li et al.99 report a progressive deterioration of memory in bvAD, mainly affecting delayed recall; this is a key aspect of the differential diagnosis of bvAD and FTD, as its onset is early and severe in bvAD.100–103 In contrast, Ossenkoppele et al.104 suggest that bvAD and FTD do not present significant differences in memory performance.

Several studies7,43,103,105 suggest that the pronounced executive dysfunction in bvAD tends to be generalised and therefore has an impact on global cognitive performance. According to Woodward et al.,106 such instruments as the Frontal Assessment Battery may detect executive function alterations; however, this instrument is more effective for differentiating between bvAD and AD-typ than between bvAD and FTD.

Toloza-Ramírez et al.12 note that phonological fluency is severely affected in bvAD, whereas semantic fluency is preserved. These symptoms, in addition to the syntactic/grammatical alterations, are essential in differential diagnosis with AD-typ.

Shelley et al.107 suggest that episodic memory is impaired early in CBS. They also suggest that this form of memory impairment is essential to identifying this entity; in fact, Sakae et al.108 report that nearly 26% of patients with CBS present memory problems in early stages.

Several studies44,108–110 report significant impairment of visuospatial skills and the development of ideomotor apraxia. Boyd et al.111 indicate that VOSP subtests are sensitive and specific for detecting CBS, helping in differential diagnosis with AD-typ. In contrast, Constantinides et al.112 suggest that differential diagnosis of CBS should not only be based on neuropsychological testing but rather should also include cerebrospinal fluid analysis.

Language impairment is marked in CBS, and has been proposed as a biomarker for early diagnosis.113–115 Cotelli et al.116 report significant syntactic/grammatical impairment with preserved comprehension. Other studies95,117 emphasise that these deficits and verbal fluency contribute to the differential diagnosis of CBS and non-fluent/agrammatic variant PPA. Other reported symptoms are problems in phonological processing118 and action naming tasks,119 anomia,120 and semantic categorisation.121

Patients with AD-typ present moderate alterations in limbic areas122 and marked atrophy of the precuneus and posterior cingulate gyrus.123–125 The heterogeneity of brain areas affected in AD-typ presents great challenges, as the classical neuropsychological symptoms do not constitute an “exclusivity” criterion for diagnosis.12,126 From a neuropathological perspective, a common criterion for identifying AD-typ is the early development of neurofibrillary tangles in the medial temporal lobe and hippocampal atrophy.127–129

In the context of lvPPA, Rohrer et al.130 report predominantly perisylvian and/or posterior parietal atrophy. Furthermore, the left inferior parietal lobe, left posterior temporal lobe, and temporoparietal junction are key areas that are vulnerable to neurodegeneration in lvPPA.131,132 Tau PET imaging suggests that the pattern of brain atrophy in lvPPA is not symmetrical.133 Another key aspect is atrophy of the hippocampus and anterior temporal lobe, which have been proposed as key findings in the differential diagnosis of lvPPA134,135 and frontotemporal lobar degeneration.133

PCA may hinder the diagnosis of AD, as neuropsychological symptoms and the pattern of brain atrophy overlap with such other profiles as AD-typ and lvPPA.81,94 In fact, our findings suggest that PCA presents significant atrophy of the occipital lobe, inferior and posterolateral temporal lobe, precuneus, and posterior cingulate cortex, areas involved in the diagnosis of AD-typ and lvPPA35; in contrast, Holden et al.90 argue that there are currently no specific biomarkers for the detection of PCA. Similarly, previous studies136,137 propose studying brain connections, as alterations in brain connectivity become more marked at more advanced stages of the disease, with greater white matter degeneration in the cingulate having a direct impact at the corticosubcortical level (frontostriatal connectivity). Indeed, Fredericks et al.138 note that changes to functional connectivity in PCA tend to be modulated by the participation of specific thalamic nuclei, such as the medial pulvinar nucleus.

The findings of our review suggest that the pattern of atrophy in CBS involves the superior temporal cortex, perirolandic region (motor and somatosensory cortex), temporoparietal region, hippocampus, subiculum, precuneus, and posterior cingulate. Previous studies44,108 have reported relative preservation of superior frontal regions, despite pronounced loss of volume in the occipital lobes and temporoparietal junction. Other studies112,139 report an asymmetrical pattern of hippocampal atrophy and significant hypoperfusion of the precuneus and posterior cingulate, as in AD-typ.

Finally, comparative studies of bvAD and AD-typ suggest that the former involves marked frontoparietal atrophy with relative preservation of medial temporal regions and subtle parietal atrophy.103,140 Stopford et al.141 suggest that the consideration of frontal hypometabolism as a biomarker in bvAD can lead to misdiagnosis as frontotemporal degeneration. Several other articles100,103,142,143 suggest that atrophy of the posterior cingulate, precuneus, hippocampus, medial temporal lobe, and anterior and posterior frontal regions should be analysed with caution due to the similarity with the patterns of atrophy reported in behavioural variant FTD.

Current diagnostic criteria for typical and atypical AD focus on both molecular and neuroimaging evidence,144,145 based on such techniques as MRI, functional MRI (fMRI), and PET.

Graff-Radford et al.146 note that MRI achieves early identification of atrophy in AD-typ in medial and lateral temporal regions and in the parietal lobe. Furthermore, they report left temporoparietal atrophy in lvPPA, parieto-occipital atrophy in PCA, frontoparietal atrophy in bvAD, and pronounced atrophy of the left motor cortex in CBS. However, Toloza-Ramírez et al.12 note that while MRI identifies this atrophy in advanced stages, early identification of these profiles requires advanced neuroimaging techniques.

Functional MRI contributes to improving the precision of anatomical localisation of brain areas involved in cognitive/linguistic processing in typical and atypical AD by analysing changes in brain metabolism.147,148 Other lines of research149–152 focus on the role of fMRI in understanding AD-typ; however, the authors stress that their findings should be interpreted with caution due to the atypical profiles of AD.

PET imaging has also been used for the early identification of typical and atypical AD, detecting important biomarkers such as the tau and amyloid proteins.153 Tau-PET imaging is able to detect tau deposition in anatomical areas associated with atrophy in typical and atypical AD, and is sensitive in distinguishing different profiles. Amyloid PET imaging identifies amyloid deposition in AD-typ, especially in the neocortex and posteromedial cortex, as well as in the medial temporal cortex, sensorimotor cortex, and visual cortex. However, despite its increasingly widespread use, results should be analysed with caution as typical and atypical variants of AD tend to display similar patterns of amyloid deposition.154–156

It should be stressed that neuroscience research has increasingly focused on observing how the brain functions, using metabolic and haemodynamic tools (e.g., fMRI). In a study using dynamic causal modelling, Friston et al.157 stress the central role of functional integration in the study of disease, which should seek to understand not only isolated brain regions, but also the connections between them. Therefore, future lines of research should focus on studying patterns of effective connectivity in language and cognition, in both typical and atypical forms of AD.

This study does not present a meta-analysis of the articles reviewed; rather, we sought to provide a qualitative analysis. The literature search was limited to 4 databases. Future studies should expand the search to avoid bias in the selection and inclusion of articles. Finally, it should be noted that the age range of the study population was 60 years and older; however, current research suggests that early-onset AD should also be taken into account.158,159

The information presented in this review is relevant to both the clinical and research settings, summarising the neuropsychological symptoms and neural correlates of AD-typ, lvPPA, PCA, bvAD, and CBS. Our descriptions encourage caution in the diagnostic workup of these atypical profiles due to the considerable neuropsychological similarities between them. Visuospatial skills are proposed as a key domain for distinguishing between PCA and CBS. Language deficits are helpful in distinguishing lvPPA from CBS, whereas difficulties with working memory are vital in discriminating between bvAD and lvPPA. Neuropsychological assessment complemented with advanced neuroimaging studies is an essential approach in reaching an accurate diagnosis, thus improving treatment programmes for these entities.

Worldwide demographic changes and the increased prevalence of AD dementia constitute a challenge for proper diagnostic workup, given the heterogeneity of atypical variants. This review is intended to serve as a guideline for healthcare professionals involved in the management of patients with neurodegenerative diseases, as it provides a detailed neuropsychological and neuroanatomical profile. It should be noted that, to our knowledge, this review is the first to jointly address lvPPA, PCA, bvAD, and CBS, and to compare them to AD-typ. Another strength is that we offer important guidelines for improving the differential diagnosis of typical and atypical AD in clinical practice, as well as some suggestions on differentiation from other recurrent neurodegenerative disorders, such as FTD and non-fluent/agrammatic variant PPA.