Early-onset dementias: diagnostic and etiological considerations

This paper summarizes the body of literature about early-onset dementia (EOD) that led to recommendations from the Fourth Canadian Consensus Conference on the Diagnosis and Treatment of Dementia. A broader differential diagnosis is required for EOD compared with late-onset dementia. Delays in diagnosis are common, and the social impact of EOD requires special care teams. The etiologies underlying EOD syndromes should take into account family history and comorbid diseases, such as cerebrovascular risk factors, that may influence the clinical presentation and age at onset. For example, although many EODs are more likely to have Mendelian genetic and/or metabolic causes, the presence of comorbidities may drive the individual at risk for late-onset dementia to manifest the symptoms at an earlier age, which contributes further to the observed heterogeneity and may confound diagnostic investigation. A personalized medicine approach to diagnosis should therefore be considered depending on the age at onset, clinical presentation, and comorbidities. Genetic counseling and testing as well as specialized biochemical screening are often required, especially in those under the age of 40 and in those with a family history of autosomal dominant or recessive disease. Novel treatments in the drug development pipeline for EOD, such as genetic forms of Alzheimer's disease, should target the specific pathogenic cascade implicated by the mutation or biochemical defect.

from early adulthood up to age 65. As such, it is critical that clinicians employ a consistent diagnostic approach when dealing with patients suff ering from EODs. Th is will maximize the chances of making an accurate diagnosis, which will facilitate discussions with the patient and their family regarding prognosis, and, in rare instances, may allow for targeted therapies that modify the course of the underlying disease; for example, chelating therapy for Wilson's disease. Th is will also allow for the more eff ective use of symptomatic therapies, including both pharmacological and nonpharmacological ones, resulting in improved quality of life through avoidance or reduction of disease-associated or drugasso ciated complications. Achieving an accurate diagnosis will also facilitate referral to clinical geneticists as this has implications beyond the affl icted patient; family members may also be at risk of developing the disease or transmitting the disease-causing mutation to their off spring, and this may aff ect family planning and insurance-related and career-related decisions. Furthermore, with the exclusion of known pathologies using a systematic approach, it may also facilitate the identification of novel genetic diseases through inclusion of patients and their extended families in research studies.
In this article, we will contrast the complex etiologies for the dementias observed across the lifespan by providing a few examples of genetic risk factors and causative mutations in the common late-onset and less common early-onset neurodegenerative pathologies, respect ively. We will also highlight the diffi culties encountered in the diagnosis and management of patients with EOD, provide diagnostic approaches based on clinical, imaging, genetic and biochemical screening, and make several recommendations approved by the CCCDTD4.

Epidemiology of dementia across the lifespan
Most population-based studies on the prevalence of dementia have focused on patients who are over the age of 60. For example, the World Health Organization has estimated that the age-standardized prevalence of dementia tends to vary between 5 and 7% of the world's population aged 60 years and older [7]. Overall, very few population-based studies have been completed on the epidemiology of EOD. Of the studies completed to date, Harvey and colleagues estimated that the prevalence of dementia in those aged 30 to 64 was 54 per 100,000 people in the United Kingdom [8]. Similarly, Ikejima and colleagues found a prevalence rate of 42.3 per 100,000 individuals aged 20 to 64 in Japan [9]. Furthermore, Th e World Alzheimer Report has estimated that between 2 and 10% of all cases of dementia begin before the age of 65 [1]. For both EOD as well as late-onset forms, the consensus is that prevalence rates increase signifi cantly with age, almost doubling every 5 years from 9/100,000 at age 30 to 156/100,000 at age 60 to 64 [7]. Numerically, the most important causes of dementia are degenerative in nature at all ages, but reversible causes are relatively more prevalent in the young [6]. Figure 1 provides a schematic representation of prevalence rates of dementia based on age ranges.

Late-onset versus early-onset dementia syndromes: etiological and therapeutic considerations
Late-onset dementia is, for the most part, considered sporadic in nature with complex genetic and environmental (mostly unknown) risk factors. Genetic analysis of complex diseases may be complicated by many factors such as incomplete penetrance, multiple disease susceptibility loci, gene-environment interactions, and diagnostic uncertainties [10]. Recently, through the application of genome-wide association studies in very large, multicenter case-control cohorts, several gene risk variants for AD and Parkinson's disease have been identifi ed, each of which confers an incremental risk to individuals possessing the risk alleles [11][12][13][14][15]. Th e genes include ones already known to be involved in the underlying pathology of these disorders -for example, apolipoprotein E (APOE) for AD and alpha-synuclein (SNCA) for Parkinson's disease -as well as many novel genes. Th e exact role that these genes play and the regional eff ects that their variants have on the brain and secondarily on neurological and cognitive functions are mostly unknown. Strategies such as reverse phenotyping, whereby these risk alleles are associated with endophenotypes or biomarkers of disease such as neuroimaging, cerebrospinal fl uid (CSF) proteomic and neuropsychological measures, are becoming increasingly important in the quest to dissect the mechanisms underlying the heterogeneity of the neurodegenerations [16,17].
Th e current prevailing hypothesis is that the risk of developing a late-onset dementia, such as sporadic AD, is due to a weak genetic eff ect of each genome-wide association study-identifi ed gene polymorphism that together across all genes synergizes to cause the symptoms and neurological decline. However, due to the nature of genome-wide association studies -specifi cally that there is clinical heterogeneity in the patients included -we hypothesize that it is more likely that each gene polymorphism or smaller subsets of the polymorphisms together with cerebrovascular risk factors (hypertension, diabetes, hypercholesterolemia, smoking, and obesity), breakdown of neuronal cellular repair mechanisms due to aging, and unknown environmental risks are responsible for several diff erent pathogenic cascades to the fi nal pathology of late-onset dementia. Th is hypothesis is illustrated in Figure 2.
In contrast, EODs are more likely to run in families and be inherited in a Mendelian fashion. As such, they are more likely to arise from single gene defects, each of which carries a large eff ect in causing the early-onset presentation. Comorbidities such as cerebrovascular risk factors are rarer in the young and are less likely to contribute to the pathogenic cascade in comparison with late-onset dementias (see Figure 1). As a result, there is a higher probability that there will be a single (or only a few) pathogenic route(s) to the EOD syndrome (see Figure 2).
Th ese important pathogenic diff erences between lateonset dementia and EODs have signifi cant future therapeutic implications. Specifi cally, we anticipate that EODs with a hypothetical single (or few) route(s) of pathogenesis may be better targeted by disease-modifying drugs that are directed against the pathogenic cascade; for example, mAbs targeted against amyloid in early-onset familial Alzheimer's disease (EOFAD). In other words, with respect to pharmacotherapy a onesize-fi ts-all disease-based approach might be feasible, although there will probably be inter-individual diff erences in response due to genetic factors that infl uence the specifi c pharmacokinetic and pharmacodynamic properties of the drug, as well as disease characteristics such as AAO, and rates of clinical progression (see Figure 2) [18]. On the contrary, if the hypothesis relating to multiple pathogenic routes for the development of late-onset dementia is correct, then we anticipate that several diff erent drugs will need to be targeted towards the diff erent pathogenic cascades involved in the fi nal common pathway to dementia in addition to aggressive management of underlying cerebrovascular risk factors; that is, a per sona lized medicine/disease-based approach (see Figure 2). We strongly believe that future research eff orts will need to focus on disentangling the likely hetero geneous mechanisms for late-onset dementia before eff ective therapies will be demonstrated.
Th e arbitrarily determined chronological age cutoff values described in the Introduction that defi ne EOD and late-onset dementia can be helpful in the diagnostic work-up because the younger the age of onset, the more likely the cause will be strongly genetic in origin and not infrequently due to familial neurodegenerative or neurometabolic disorders (see Figure 1) [19]. In con sidering these age cutoff values, however, biological age rather than chronological age must also be taken into consideration. For example, a patient aged 50 years with several uncontrolled cerebro vascular risk factors may have a more similar pathogenic cascade leading to dementia to someone who develops dementia in their 70s, in comparison with someone who develops it at the age of 50 but does not have any signifi cant medical comorbidities. Th ere may be variable expressivity of the age of onset depending on the strength of the contri buting genetic eff ects and the presence of comorbidities leading to clinical syndromes that overlap in terms of the age of onset, but may indeed be due to diff erent etiologies as shown in Figure 1. We will now discuss the most common types of EODs, reviewing clinical, genetic, and pathological features.

Early-onset familial Alzheimer's disease
EOFAD is the most well-known group of persons with EOD because of the autosomal dominant pattern of Figure 1. Prevalence rates of early-onset dementia and relative contribution from presumed etiological mechanisms. Schematic representation of the prevalence rates for early-onset dementia and the relative contribution from presumed etiological mechanisms in three diff erent age ranges. Areas of circles represent the estimated prevalence rates within the specifi ed age grouping.
inheritance aff ecting fi rst-degree relatives, the high penetrance of the three known major mutations, which are examined for in genetics clinics, and the recent interest in the progression of biomarkers through the asymptomatic, mild cognitive impairment and early dementia stages, studied as part of the Dominantly Inherited Alzheimer Network and the Alzheimer's Preven tion Initiative [20][21][22]. Hopes are that studies such as this may lead to anti-amyloid and other drug trials in this group of EOFAD patients, who traditionally have been excluded from clinical trials.
EOFAD has been extensively reviewed by Wu and colleagues, the highlights of which are as follows [23]. EOFAD is a condition that represents up to 5% of all AD cases in clinical practice. To date, 230 mutations in presenilin (PS1, PS2) and amyloid precursor protein (APP) genes have been identifi ed in EOFAD. Th e mutations within these three genes (PS1, PS2, APP) aff ect a common pathogenic pathway in APP synthesis and pro teolysis, which leads to excessive production of amyloid beta. More recently, using exome sequencing, mutations in the Sortilin-related receptor LR11/SorLA (SORL1) were discovered in autosomal dominant, earlyonset AD index cases, who did not harbor mutations in the other AD-causing genes [24]. Initial interest in this gene was identifi ed through a large case-control genetic association study of markers within SORL1 [25], and the more recent discovery might suggest that a few individuals included in this study harbored mutations that were in linkage disequilibrium with the studied markers.
For the most part, the clinical presentation of EOFAD is similar to that of sporadic AD, with an amnestic syndrome being common to both in the majority of cases although atypical AD variants are more common in early-onset cases [26]. However, there are some distinctive features including early AAO, positive family history, and a variety of noncognitive neurological symptoms and signs. Despite relatively similar biochemical defects, there is marked phenotypic heterogeneity among diff erent mutations of EOFAD. Clinical symptoms start at an earlier age for carriers of PS1 mutations in comparison with those with PS2 or APP mutations. Studies in pre symptomatic mutation carriers reveal amyloid patho logical biomarker abnormalities including positive uptake of Pittsburgh Compound B on positron emission tomo graphy (PET) and lowering of amyloid beta in CSF at least 10 years before symptoms emerge [20].

Frontotemporal lobar degeneration spectrum disorders
In Canada, FTLD accounts for about 12% of referrals in those under age 70 to tertiary dementia clinics [27]. FTLD comprises a heterogeneous group of neuro degenera tive pathologies that cause clinical dementia syndromes characterized by prominent decline in social conduct and behavior, including, for example, disinhi bition, apathy, and hyperorality, and/or prominent language dysfunction, referred to as behavioral variant and primary progressive aphasia (PPA) variants of frontotemporal dementia, respectively [28,29]. As specifi ed in the name, symptoms result from focal cortical and subcortical atrophy involving the frontal and temporal lobes, most often asymmetrically in the early disease stages with prominent right-sided involvement presenting as behavioral variant frontotemporal dementia and left-sided involvement as PPA. However, the neurodegenerative pathology relentlessly progresses to involve both sides and, therefore, disease evolution to mixed forms is the rule rather than the exception [30,31]. FTLD represents the second most common form of EOD after AD with AAO ranging from 45 to 64 years [32].
Familial inheritance of FTLD is even stronger than for AD, representing up to 40% of all FTLD encountered clinically [33]. Within FTLD, heritability can also vary substantially between the diff erent clinical syndromes observed [34]. To date, three genes have been identifi ed to be responsible for the majority of autosomal dominant forms of FTLD: microtubule-associated protein tau (MAPT), progranulin (GRN), and the C9ORF72 gene [35][36][37]. Mutations in the valosin-containing protein (VCP), chromatin modifying protein 2B (CHMP2B), trans active DNA-binding protein (TARDBP) and fusedin sarcoma (FUS) have also been identifi ed as rare causes of familial FTLD although most cases related to FUS are sporadic in nature [38]. In genetic FTLD, the reported range of age of onset is highly variable from 29 to 81 years, particularly in those with GRN or C9ORF72 mutations [39,40]. Findings of motoneuron disease may coexist with FTLD and have rarely been associated with GRN [41], TARDBP and CHMP2B mutation, but are commonly seen with C9ORF72 repeat expansions [33].
Th e discovery of the various FTLD-causing mutations has largely been driven by the genomic analysis of subgroups of patients classifi ed according to the hetero geneous pathologies underlying FTLD. Th ese can be largely divided into two broad subgroups: Tau-positive cases (can be associated with MAPT mutations), and Taunegative cases with ubiquitin and TDP43 inclusions (FTLD-TDP) can be associated with GRN and VCP muta tions, and C9ORF72 hexanucleotide repeat expansions [33,38]. Some cases of FTLD with Tau-negative, ubiquitin-positive inclusions may not demonstrate any TDP43 immunoreactivity, and FUS can be the pathological protein associated with FUS mutations [42]. Finally, there are cases of ubiquitin-positive FTLD without TDP43 and FUS immunoreactivity, and some of these cases can be associated with CHMP2B mutations [33].
Corticobasal syndrome (CBS) and progressive supranuclear palsy (PSP) are atypical Parkinsonian syndromes that occur along the spectrum of FTLD due to overlapping pathologies and clinical presentations [43]. CBS presents with the combination of various cortical and extrapyramidal features usually with striking asymmetry [44]. Th e cortical features may include focal or asymmetric ideomotor apraxia, alien limb phenomenon, cortical sensory loss, visual or sensory hemi-neglect, constructional apraxia, and focal or asymmetric myoclonus [44]. Th e extrapyramidal features may consist of appendicular rigidity lacking prominent and sustained levodopa response, and appendicular dystonia [44]. An early dementia onset may be the most common initial presentation of CBS [45], and this often has overlapping features with FTLD such as a progressive nonfl uent aphasia/apraxia of speech or a behavioral presentation [46]. PSP is characterized by progressive axial greater than appendicular rigidity, early postural instability with falls, dysphagia, and vertical supranuclear gaze palsy, which is the hallmark feature of this disease although it may often not emerge until late in the course [47]. Patients may also present with frontal behavioral abnormalities and executive cognitive decline as well as aphasia [47]. Although the clinical presentation of PSP is most probably due to an underlying Tauopathy, in particular PSP pathology [48], clinically defi ned CBS may be due to Tau pathologies (for example, corticobasal degeneration (CBD), Pick's disease, PSP), non-Tau pathologies (for example, FTLD-TDP) and mixed amyloid and Tau pathology (for example, AD), making it more pathologically heterogeneous [46]. CBS typically presents in the sixth to eighth decade of life and has a mean age of onset of approximately 63 years (standard deviation 7.7 years) [49], similar to that observed in PSP [50].
Genetic association studies of sporadic cases of CBD and PSP with the typical age of onset demonstrate that the MAPT H1 haplotype is over-represented in cases compared with controls [51,52]. However, since this is a common haplotype, it is probably not that useful to distinguish amongst the various atypical parkinsonian syndromes, although one small study (from the genetic standpoint) demonstrated a nonsignifi cantly higher odds ratio of the H1/H1 genotype in the Richardson variant of PSP than in the PSP-Parkinsonism variant [47]. In the rare event that CBS and/or CBD pathology are observed to segregate in a family, other members are typically aff ected with FTLD and/or PSP demonstrating overlap in these conditions [53][54][55][56][57][58][59][60]. Tau pathology has been most commonly observed with CBD or PSP inclusions [56], with a few families demonstrating MAPT mutations [56]. Some families have demonstrated non-Tau pathology as well [53,55]. Mutations in GRN have been shown to segregate with CBS phenotypes in family members due to FTLD-Ubiquitin pathology [31]. Familial cases of CBS and PSP may have an earlier AAO than that of sporadic cases.

Lewy body spectrum disorders: Parkinson's dementia and Lewy body dementia
Parkinson's disease dementia (PDD) and dementia with Lewy bodies (DLB) demonstrate varying combinations of extrapyramidal, cognitive, neuropsychiatric, and autonomic symptoms. Th e distinguishing clinical feature between these disorders arbitrarily relates to the time of onset of motor symptoms. PDD presents with early parkinsonism (tremor, rigidity, bradykinesia) that responds well to dopaminergic therapy, followed by cognitive and neuropsychiatric decline at least 1 year after the motor onset [61]. In contrast, DLB presents with an early dementia consisting of fl uctuations in attention and alertness as well as visual hallucinations [62]. Parkin sonism may be contemporary with the dementia or may develop later [62].
Clinical symptoms and signs of these diseases occur as a result of profound neurotransmitter dysfunction includ ing dopaminergic systems (motor and executive dysfunction), serotonergic systems (mood and halluci nations), noradrenergic systems (gait, mood, and inattention), and cholinergic systems (hallucinations, inattention, reductions in semantic fl uency, executive and visuospatial dysfunction) [63,64]. Th e neurotransmitter dysfunction occurs secondarily to the presence of brain stem, cortical and limbic Lewy bodies and Lewy neurites, which are comprised of patho logical intraneuronal aggregates of alpha-synuclein [62]. Th ere is increasing evidence that beta-amyloid plaques may also contribute to the dementia seen in PDD [65][66][67] and DLB [68]. As a result of this clinical and neuropathological overlap, DLB and PDD are considered disorders that occur along the spectrum of Lewy body disease and are classifi ed as synucleinopathies.
Similar to AD, Lewy body spectrum disorders are typically of late onset and are, for the most part, considered sporadic in nature with complex genetic and unknown environmental factors that increase risk for disease. In a prospective study, PD patients with the MAPT H1/H1 genotype had a signifi cantly increased risk of developing dementia (odds ratio = 12.1) and a more rapid rate of cognitive decline over the longitudinal assessment period [69]. A second prospective study did not fi nd an association between change in Mattis Dementia Rating Scale version 2 scores over time and MAPT H1 haplotype in PD [70]. Th e inconsistent results for the MAPT H1 haplotype are probably due to small sample sizes in both studies; however, there remains the possibility that misdiagnosis of CBD or PSP patients as PDD or alternatively mixed Tau and Lewy body pathologies may be contributing to the discrepant results. A more recent pathological study has demonstrated an increased frequency of the APOE ε4 allele in DLB with AD pathological changes, pure DLB and PDD compared with controls, implicating a role for the ε4 allele in lateonset sporadic forms of Lewy body spectrum disorders [71]. Th is further supports an overlap between AD and Lewy body spectrum disorders.
Although rarer than AD and frontotemporal dementia in patients younger than 65 years, early-onset Lewy body spectrum disorders have been described. Several earlyonset families segregating phenotypes of DLB and Parkinson's disease have been reported due to SNCA triplications [72] or the SNCA missense mutation, p.A53T [73]. A number of other early-onset families with Lewy body spectrum disorders have been identifi ed, but a causative gene has yet to be identifi ed [74]. In further support of the link between AD and Lewy body patho logy, PS1, PS2, and APP gene mutations can cause a DLB phenotype due to mixed Lewy body and AD pathology [75][76][77][78].
More recent evidence is emerging for a role of heterozygous glucocerebrosidase (GBA1) mutations in Lewy body spectrum disorders [79][80][81]. A multicenter case-control study of 721 cases clinically diagnosed with DLB, several with pathological confi rmation, and 151 cases of PDD demonstrated an increased risk of DLB and PDD in GBA1 mutation carriers compared with controls [82]. GBA1 mutation carriers had an earlier age of onset (63.5 years) than noncarriers (68.9 years) and also had higher Hoehn and Yahr stage disease and Unifi ed Parkinson's Disease Rating Scale scores [82]. Lyzosomal dysfunction may therefore be involved in the pathogenic cascade leading to alpha-synuclein aggregation in the subgroup of patients with GBA1 mutations, although the specifi c mechanisms are not yet fully elucidated [83].

Vascular cognitive impairment and mixed Alzheimer's with cerebral small vessel disease
Mixed Alzheimer's with subcortical ischemic white matter disease (that is, small vessel disease (SVD)) is the most common type of late-onset dementia observed [84,85] and is also probably responsible for a good proportion of early-onset cases in the later age ranges (for example, 50 to 65 years) because of the high prevalence of cerebrovascular risk factors in the general population [86]. Patients with this mixed dementia state often have signifi cant impairments in episodic memory typical of AD, but additionally demonstrate impairments in executive functions, psychomotor speed of processing, and mental fl exibility (that is, frontal-subcortical dementia) [87][88][89].
Pure vascular cognitive impairment is comparatively uncommon in contrast to mixed AD with SVD, and usually presents with an insidious decline in cognition and a frontal-subcortical dementia with memory benefi tting from cueing [90,91]. A stepwise trajectory of decline may less commonly be observed with vascular cognitive impairment [90,91]. Together with signifi cant periventricular frontal white matter SVD burden and associated periventricular atrophy, an apraxia of gait may also be observed and this is often referred to as lower limb or vascular parkinsonism. Th is resembles the gait disorder observed in normal pressure hydrocephalus. Although the latter condition needs to be considered at diff erential diagnosis, vascular cognitive impairment/parkinsonism is often misdiagnosed as normal pressure hydrocephalus and a therapeutic lumbar puncture followed by gait assessment is required to diff erentiate the two entities.
Cerebrovascular disorders are probably the most heterogeneous of the pathologies that can cause dementia in the young. Th e reason for this heterogeneity is because of the large number of diseases that can result in overt strategic infarction (hemorrhagic or ischemic) and/or covert subcortical ischemic white matter disease (SVD) [92]. Discussing all of the stroke etiologies that can lead to young-onset vascular cognitive impairment is beyond the scope of this review. However, a few notable examples will be discussed briefl y, including cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), cerebral vasculitis, and cerebral amyloid angiopathy. CADASIL is the most common genetic cause of young-onset vascular cognitive impairment and is associated with migraine, recurrent strokes, dementia, and often psychiatric features including psychosis [93]. Diagnosis is suggested through the appropriate history including that of the family, examination fi ndings, as well as magnetic resonance imaging (MRI) showing lacunar-type infarctions and SVD that can be confl uent as the disease progresses and that also involves the anterior temporal lobes (see Figure 3a) [93]. Th e diagnosis is confi rmed through skin biopsy showing granular deposits in the basal lamina of small vessels and with the identifi cation of NOTCH3 gene mutations [93].
Primary cerebral (CNS) vasculitis or angiitis is typically associated with headache, progressive encephalopathy, and multiple strokes [92,94]. Non in vasive and invasive cerebral angiography techniques can show the classic vessel beading pattern dependent on the size of the vessels involved (that is, medium-size to large-size vessels), and CSF often demonstrates a lymphocytic pleiocytosis and elevated protein levels [94]. In diagnostically challenging cases, brain biopsy remains the gold standard [94]. Histologically, a lymphocytic infl ammatory process consisting of plasma cells, histiocytes, neutrophils, and eosinophils is most commonly seen, while the classic segmental granulomatous vasculitis with multinucleated giant cells is less often observed [94]. Secondary causes of vasculitis such as those associated with connective tissue diseases and infections must be excluded [94]. Th e utility of the various tests used to diagnose primary CNS vasculitis has been studied and, overall, the sensitivity, specifi city, as well as positive and negative predictive values of each individual test are generally low [95]. Accurate diagnosis should therefore rely on the appropriate clinical history, examination fi ndings, and correlation with multiple diagnostic investigations [95]. Finally, cerebral amyloid angiopathy can cause cerebral SVD and lobar micro and frank hemorrhages, and these pathologies can contribute to cognitive decline in patients with and without Alzheimer's pathology [85]. Later onset cases of cerebral amyloid angiopathy have been associated with APOE ε4 homozygotes [96], while earlier onset cerebral amyloid angiopathy cases with AD have been associated with duplications in the APP gene [97].

Other causes of early-onset dementias
A number of rare metabolic genetic conditions can also present as young-onset dementia. Providing a detailed account of these disorders is beyond the scope of this review, and an approach to investigating them has been provided by others [19]. However, a selected list of metabolic/genetic disorders that can present with progressive dementia in the young, organized by the most common neurological features observed, is summarized in Table 1. Th e table also presents the genetic/ metabolic defects, and other associated clinical features of these disorders, which may assist in the diagnostic work-up. Th e majority of these demonstrate autosomal recessive, X-linked, or mito chon drial inheri tance. Ascertain ment of a family history of consanguinity or marriage within a genetically homoge neous group -for example, Ashkenazi Jews -and absence of disease in a generation suggest autosomal recessive inheritance. If males are prominently, or more severely, aff ected than females, an X-linked inheritance is likely, while maternal lineage supports mitochondrial disease. Th ese rules of thumb may assist in screening for the appropriate genes and biochemical testing. Although these metabolic genetic diseases are relatively rare, their prevalence is higher in those under the age of 35 and therefore appropriate genetic testing and biochemical screening should be conducted based on the clinical presentation in aff ected individuals under this age (Table 1) [4]. Accurate diagnosis will allow proper genetic counseling and provide prognostic guidance to the family, which is an integral component in the manage ment of dementia. Th ese include disorders of amino acid and organic acid metabolism, lysosomal storage diseases, leukodystro phies, mitochondrial diseases, and disorders of metal meta bolism. While many of these genetic meta bolic diseases are multisystem disorders, most will also suff er a signifi cant degree of cognitive impairment [98]. In some cases, such as Niemann-Pick type C, dementia can be the main and only clinical presentation.
In addition to genetic-based neurodegenerative diseases, other systemic diseases can also lead to progressive cognitive decline that mimics EOD. Some reviews, such as that of Fadil and colleagues [99], list as many as 61 causes of EODs. For example, infectious diseases such as HIV, Lyme and syphilis, autoimmune infl ammatory diseases such as lupus with CNS involvement, paraneoplastic syndromes with limbic en cepha litis, and sleep disorders such as obstructive sleep apnea can all lead to progressive cognitive impairment ( Table 2). Th e autoimmune encephalopathies are another group of clinically heterogeneous disorders associated with cognitive impairment in the young and, although they can be paraneoplastic in nature, they are most often not associated with tumors [100]. Antibodies to compo nents of N-methyl-d-aspartate receptors, 2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid receptors, gammaaminobutyric acid-type B receptors, glycine receptors, and voltage-gated potassium channel complexes can be detected in aff ected patients, and symptoms are responsive to treatment with immunomodulatory drugs [100]. Th is group of disorders therefore represents a treatable cause of encephalopathy in the young and should not be missed.

Diagnostic process and genetic testing
Since the diff erential diagnosis is wide, a systematic clinical and investigational approach is needed to ensure all potentially treatable conditions have been ruled out and to identify associated clinical features that may help to narrow down the diff erential diagnosis; that is, dementia-plus disorders [6]. Th e examination should  Electro encephalogram and CSF analysis is recom mended in the majority of patients with EOD to rule out epileptic, infl ammatory, and infectious causes and to further refi ne the diff erential diagnosis. More specialized biochemical/ genetic testing for neurometabolic disorders will be dependent on AAO, family history, and associated clinical features (Table 1). Th ere is also evidence from a large, retrospective case series that brain biopsy is potentially useful when diagnosis cannot be made by standard investigations and may reveal treatable diagnoses in about 10% of cases [101]. We also strongly recommend longitudinal follow-up, because in the very early stages of all EOD pathologies there may be considerable overlap and not all clinical features will have necessarily emerged. Diagnostic specifi city will therefore improve with subsequent assessments in the early to mid stages of the disease process. Cognitive screening and, preferably, neuropsychological testing should be completed to fully understand the cognitive profi le involved. Neuropsychometric testing is also helpful in tracking disease progression and, potentially, response to treatment. Cognitive impairment can be broadly classifi ed into two major subclasses based on the brain regions involved. 'Subcortical dementia' results  from injury to the white matter pathways and damage to the basal ganglia and thalamic structures with psychomotor slowing, frontal dysexecutive syndrome, and memory retrieval diffi culties. Alternatively, cortical dementia is characterized by problems in specifi c cognitive domains (for example, Alzheimer disease with episodic memory impairment) with relative preservation of processing speed. We will now outline clinical pearls and investigational approaches used in specialist memory and movement disorder clinics in Canada to assist with the diagnostic work-up of EOD. Th e approaches will focus mainly on slowly progressive forms of dementia (for example, >12 months) and not on rapidly progressive forms, such as Creutzfeldt-Jakob disease, which have been discussed elsewhere [102]. Th e profi le of cognitive defi cits based on history and observed on neuropsychometric testing in the early stages of young-onset dementia will serve as the initial diagnostic starting point and will now be expanded upon based on the presence of associated clinical features, neuro imaging, and laboratory investigations.

Early-onset dementias/syndromes characterized by predominant memory defi cits
Most patients with cognitive diffi culties (or their families) will complain of poor memory. However, when a patient with insidious EOD has amnesia as the primary complaint and this is corroborated on episodic memory testing, which demonstrates no or minimal benefi t from cueing, then early-onset AD should be considered high on the diff erential diagnosis (Figure 4). Often there are no other associated clinical features and their absence may support a typical Alzheimer's presentation [26]. Th e MRI and functional imaging signature of typical AD, reviewed by McKhann and colleagues [103], is summarized in Figure 4. Figure 5b shows mesiotemporal atrophy on MRI of a patient with EOFAD due to PS1 mutation. Routine CSF studies should be normal, but specialized research testing of CSF demonstrates reduced amyloid-beta levels and increased total-tau and phospho-tau levels [103].
Research PET scans may be obtained that demonstrate retention of ligands specifi c for amyloid, such as Pittsburgh Compound B, indicative of increased fi brillar  amyloid brain accumulation [103]. Th e research CSF and PET studies are not currently approved in Canada for the investigation and diagnosis of early-onset or late-onset AD and should be considered mainly for diagnostically challenging cases depending on availability and funding. Th e presence of a family history suggesting autosomal dominant AD should prompt mutation screening for PS1, PS2, APP, and SORL1 in that order (Figure 4). Even in the absence of a strong family history, AD mutation screening should be strongly considered especially when the age of onset is very early and there are no associated medical comorbidities. If there is a history of alcoholism or severe malnutrition, then Wernicke-Korsakoff syndrome or alcohol-related dementia should also be considered on the diff erential diagnosis although confabulation is commonly observed in this condition and is not typical of AD (Figure 4). Furthermore, ophthalmoplegia, ataxia, and signs of chronic liver disease may be present and therefore should be looked for on examination. MRI imaging in Wernicke-Korsakoff syndrome shows several characteristic features that distinguish it from AD ( Figure 4) [104]. Ataxia has also been observed in some cases of EOFAD associated with PS1 mutations, and genetic testing should be considered for this in the absence of an alcohol or malnutrition history (referred to as AD plus in Figure 4) [23].
EOFAD may also be associated with several other clinical features (referred to as AD plus in Figure 4) [23]. If there is a history of seizures then PS1 and PS2 mutation screening should be considered once other diagnoses such as temporal lobe epilepsy and limbic encephalitis due to paraneoplastic, autoimmune, or viral etiologies have been excluded (Figure 4). In mesiotemporal lobe epilepsy, there will often be a history suggestive of simple or complex partial seizures and symptoms of memory loss may occur in association with or be aggravated by ictal events [105,106]. Limbic encephalitis may present with episodic memory disturbance, seizures, confusion, and psychiatric disturbances [107]. A subacute course is typical for paraneoplastic-mediated and autoimmune-mediated etiologies (that is, autoimmune encephalo pathies) with several autoantibodies identifi ed, and a more acute course with fever is consistent with viral causes, such as herpes simplex, which should not be missed and always investigated with CSF sampling (Figure 4) and covered empirically with anti-viral therapy, such as acyclovir [108].
Early-onset AD with movement disorders such as parkinsonism or myoclonus and/or psychiatric features such as hallucinations and psychosis may be seen with PS1 and PS2 mutations [23]. AD with spastic paraparesis or aphasia has been associated with PS1 mutations [23]. Finally, AD with intracerebral hemorrhage has been observed with APP mutations or duplications [23]. Appropriate mutation screening for EOFAD genes may therefore be guided by these clinical associations referred to as AD plus in Figure 4.

Early-onset dementias/syndromes characterized by cognitive defi cits in other domains
Th e approach to EODs with defi cits in other cognitive domains is more complex than those associated with primary memory problems. Th ere are far more diseases that fall under this category and the clinical features of these diseases can overlap signifi cantly with prominent executive dysfunction as a main fi nding in most cases, alone or in combination with other defi cits. Furthermore, there are also early-onset atypical variants of AD that fall under this highly heterogeneous group, which makes sorting out the diff erential diagnosis more challenging [26]. Since most of these EOD syndromes described in this section are due to non-Alzheimer's pathology, we direct you to the recent consensus guidelines on the diagnosis and management of disorders associated with dementia that have been proposed by the European Federation of Neurological Societies-European Neurological Society task force [109]. We will fi rst discuss the diagnostic work-up for some of the more pure nonamnestic EOD syndromes (at least in their earliest stages) and will then discuss those that present with mixed defi cits as well as other associated clinical features. Figures 6 and 7 present algorithms that can further aid in the diagnosis of nonamnestic presentations.
PPAs are clinically and pathologically heterogeneous in nature and are most likely to be due to FTLD, although variants of AD may also cause language disturbances. It is not uncommon that activities of daily living of affl icted patients may be preserved for a number of years apart from those relying on language function, although in most patients additional clinical features, such as behavioral dysregulation, executive dysfunction, apraxia and parkinsonism, can eventually emerge. Th e nonfl uent variant (progressive nonfl uent aphasia) presents with agrammatic and halting, eff ortful speech ( Figure 6) [28]. Th ere may be an associated apraxia of speech characterized by articulatory groping, dysprodic errors, and distorted speech sounds [110]. Pathologies that have been associated with mainly sporadic forms of the nonfl uent and/or apraxia of speech variant of PPA are Tau-positive FTLD, as well as CBD and PSP [28,110]. FTLD-TDP due to GRN mutations has also been observed with the nonfl uent PPA variant, although this is less common (Figure 6) [30,111]. Impaired naming to confrontation and single-word comprehension characterizes the semantic variant of PPA (semantic dementia; Figure 6) [28]. In sporadic forms FTLD-TDP pathology prevails [28], although familial cases have been associated with Tau-positive FTLD and MAPT mutations ( Figure 6) [112]. Finally, the logopenic variant of PPA (logopenic aphasia) presents with impaired repetition of sentences and phrases as well as with diffi culties in retrieving single words in spontaneous speech and naming [28]. Th e associated neuroimaging signatures of these PPA variants are summarized in Figure 6. In PPA cases, genetic testing should be guided by a positive family history of autosomal dominant disease and clinical presentation, with GRN and C9ORF72 mutation screening in nonfl uent cases and with MAPT screening in semantic variant cases ( Figure 6). PS1 and GRN mutation screening may be considered in select cases of logopenic aphasia ( Figure 6) [23].
If attention, working memory, and executive function defi cits are prominent, then the diagnostic approach is strongly guided by associated clinical features and structural and functional imaging. If behavioral features such as disinhibition, apathy, personality change, loss of self-regulation, and social impropriety are observed with executive dysfunction, and if MRI shows right frontal and anterior temporal atrophy worse than on the left (Figure 5a) with similar regions showing reduced activity on SPECT/PET, then behavioral variant frontotemporal dementia is high on the diff erential diagnosis (Figure 7). Appropriate genetic testing for GRN and MAPT mutations as well as hexa nucleotide repeat expansions in C9ORF72 should be performed, especially if there is a strong autosomal domi nant family history (Figure 7) [113]. Th e order of FTLD genetic testing should be guided by the clinical and familial history, examination, and MRI fi ndings. GRN mutations are associated with more markedly asymmetric frontotemporal atrophy that also involves the parietal lobes ( Figure 5a) and cases that eventually may evolve clinical features of CBS, whereas cases with MAPT mutations tend to have less atrophy and a stronger predilection for temporal areas with more focal involvement [38]. Finally, both familial and sporadic cases with C9ORF72 repeat expansions have more variable presentations including slowly progressive variants [114] and promi nent neuropsychiatric illness [115], as well as association with motoneuron disease; atrophy tends to be more symmetrical involving the dorsolateral, medial, and orbitofrontal lobes [116]. Th e primary psychoses, such as bipolar disorder and schizophrenia, can also present with prominent behavioral features and are also considered in the diff erential diagnosis.
Vascular cognitive impairment also demonstrates prominent dysexecutive features and is suggested by the appropriate clinical history of cerebrovascular risk factors, stroke and/or focal neurological defi cits conform ing to specifi c arterial territories (Figure 7). In the absence of this clinical history and fi ndings, MRI of the brain will invariably show SVD (Figure 3b) and/or cortical infarcts if they are located within the so-called silent regions (Figure 7). If cerebrovascular risk factors are not present, then CADASIL and CNS vasculitis should be considered (Figure 7). Patients with migraine, short stature, and strokes that do not respect vascular territories should be investigated with CSF and serum studies, particularly looking for increased lactate and pyruvate. A high lactate:pyruvate ratio is consistent with mitochondrial encephalopathy, lactic acidosis and strokelike episodes; diagnosis is made through skeletal muscle biopsy demonstrating ragged red fi bers, and genetic testing should examine for mutations in the mitochondrial DNA genes, MT-TL1 (~80% of cases) or MT-ND5 (Figure 7) [117]. APP duplications or mutations should be screened for if the EOD syndrome is associated with lobar hemorrhages (that is, cerebral amyloid angiopathy).
If executive dysfunction and parkinsonism (tremor, rigidity, akinesia/bradykinesia, and/or gait disorder) is evident on the examination, then atypical Parkinsonianrelated dementias are highest on the diff erential diagnosis (Figures 6 and 7). Th e most common subtype will be disorders falling under the Lewy body spectrum. Th e presence of visual hallucinations, fl uctuations in attention and alertness, visuospatial defi cits, rapid eye movement behavioral disorder, and dysautonomia are associated clinical features common to this spectrum. Not all features are always present, which broadens the diff erential diagnosis. Th e cognitive and neuropsychiatric symptoms of DLB often, but not always, respond to cholinesterase inhibitors and the response is usually more marked than that seen in AD. PDD patients will have a more robust response to levodopa and may develop postural instability as a late feature; on-off fl uctu ations and dyskinesias may be observed. DLB usually has at least a partial response to levodopa with signifi cantly less motor complications, such as dyskinesia; postural instability may be an early feature. Dopamine agonists and amantadine typically produce increased confusion, sedation, and hallucinations so their use should be cautioned in PDD and DLB. Levodopa may also produce these symptoms, but usually to a much lesser degree. Impulse control disorders observed in PDD due to excessive dopaminergic therapy, especially with dopamine agonists, may also contribute to the behavioral and cognitive executive disorder, confounding the diagnosis. SPECT and PET studies of PDD and DLB show hypoperfusion and hypometabolism, respectively, in biparietal regions extending into the lateral occipital cortex (Figures 6, 7, and 8c). Mesiotemporal atrophy may be seen and this may suggest concomitant AD pathology. An electroencephalogram often shows generalized slowing. Reduced putaminal binding of presynaptic dopamine trans porter ligands on SPECT may also be used to help diff erentiate DLB from AD [118], but this modality is currently not available in Canada. We would only recom mend genetic testing in clinically typical Lewy body disease if there is a family history of at least two fi rst-degree relatives aff ected. SNCA triplications and muta tions may be screened. If other family members have AD phenotypes then PS1, PS2, and APP screening may be considered.
If the dysexecutive Parkinsonian syndrome is not associated with cognitive fl uctuations or hallucinations, then CBS or PSP should alternatively be considered. CBS typically demonstrates striking lateralization of clinical fi ndings. Atrophy on T1 sequences and subcortical T2/ fl uid-attenuated inversion recovery signal changes on MRI representing gliosis, as well as hypoperfusion/ hypometabolism on functional imaging, are typically worse contralateral to the most aff ected side of the body and aff ect frontoparietal areas more than temporal areas (Figures 6, 7, and 8a). If the rigidity is present in axial muscles with nuchal hyper extension and frontalis overactivity, then PSP should be considered even in the absence of the vertical supra nuclear gaze palsy. MRI imaging typically demonstrates the so-called humming bird sign evident on midline sagittal sections and is due to atrophy involving the midbrain (Figure 7 and 8d) [47]. Levodopa responsiveness is poor in CBS and PSP and, if it is present, it is usually not sustained. Niemann-Pick disease type C also often has a vertical supranuclear gaze palsy present as well as cerebellar ataxia, dysarthria, psychiatric disturbances, and splenomegaly (Table 1) [119]. Diagnostic testing of suspected cases of Niemann-Pick type C should be per formed via skin biopsy with culturing of fi broblasts and examination for intracellular cholesterol accumulation (fi lipin staining) in addition to screening for mutations in the NPC1 (95% of cases) and NPC2 (5%) genes (Table 1) [119]. Other FTLD spectrum disorders, such as FTLD with Parkinsonism-TDP or Parkinsonism-Tau linked to chromosome 17 (GRN mutations and MAPT mutations, respectively), can also be associated with atypical parkin sonism. However, there are usually associated symptoms and MRI fi ndings of FTLD observed.
Wilson's disease should be screened for in cases of age of onset <40 years. Investi gations reveal low ceruloplasmin levels, elevated 24-hour urinary copper levels, and the presence of Kayser-Fleischer rings through slitlamp examination [120]. Atypical parkinsonism with a frontal-subcortical dementia, psychosis, and a wingfl apping/beating tremor may be observed [120]. Liver function tests and abdominal ultrasound should also be carried out. MRI of the brain may show the face of the giant panda sign (Figure 8b) [120]. Diagnosis is made via liver biopsy showing copper accumulation and through demonstra tion of recessive mutations in the ATP7B gene [120]. Wilson's disease is treatable acutely with chelating therapy and chronically with maintenance zinc supplemen tation, and therefore the diagnosis should not be missed. Later onset cases of Wilson's disease have also been described so we recommend a low threshold for screening if the Parkinsonian syndrome is not consistent with some of the more common atypical Parkinsonian variants in those older than 40 years.
EODs that present with prominent visuospatial and visuoperceptual defi cits on history and neuropsycho logical testing represent another neurodegenerative syndrome termed posterior cortical atrophy (see Figure 6). Patients often present to psychiatrists and ophthalmologists before being assessed by neurologists due to the atypical presentation of this entity. Posterior cortical atrophy typically occurs between ages 50 and 65 years and is often associated with features of Bálint's syndrome (asimultagnosia, optic ataxia, and oculomotor apraxia), Gerstmann's syndrome (acalculia, right-left confusion, fi nger agnosia, and agraphia), alexia, limb apraxia (both ideational and ideomotor), and working memory defi cits [121][122][123]. MRI fi ndings are summarized in Figure 6. Fluorodeoxyglucose PET and perfusion SPECT show hypometabolism and hypoperfusion in similar regions to those aff ected on the MRI, with functional defi cits also observed in the middle and posterior cingulum, the pulvinar, as well as the middle and superior frontal gyri corresponding to the frontal eye fi elds [121,122]. Posterior cortical atrophy is most commonly associated with AD pathology; however, DLB, CBD, and prion disease have also been observed [121].
Th ere are many other exceedingly rare dementia-plus conditions that can present with early AAO and associated frontal-executive features with and without neuropsychiatric features. Table 1 describes the clinical and imaging features, biochemical defects, and genetic mutations observed in these rare dementia-plus conditions organized according to most common neurological associations.

Recommendations regarding early-onset dementia approved by the CCCDTD
• All patients with EOD should be referred to a specialist (memory or movement disorders) clinic, preferably one with access to genetic counseling and testing as well as biochemical screening when appropriate. • Th e investigation of genetic etiologies should be guided by the detailed assessment and investigations (MRI, functional imaging, laboratory investigations, and so forth) performed by the specialist neurologist in conjunction with the clinical geneticist, who will provide information to the patient and their family regarding implications of the potential fi ndings. • Th e cost of genetic counseling and testing should be covered by public funding. • If AAO is <40 years then rare genetic metabolic dis orders should always be investigated if the specifi c dementia diagnosis is not evident through other investigations. • Physicians should be sensitive to the special issues associated with EOD, in particular with regard to loss of employment and access to support services appropriate for that age group. • Considering the rarity of EOD, a national registry for interested at-risk individuals, mutation carriers and symptomatic patients will facilitate therapeutic research. • Th is registry should be supported by public funding.

Conclusions and future directions
Th e diagnosis and management of EODs is challenging but is an important component of neurologic referrals. Th is review considers the diagnostic investigations that are currently available to most neurologists, geriatricians, and geriatric psychiatrists worldwide in their clinical practice. It is fortunate that many genetic markers are emerging for AD and FTLD, and there is increasing know ledge about biomarkers that can potentially be tracked years before clinical symptoms emerge in EOFAD. Th e emerging biomarkers include fl uorodeoxyglucose PET and amyloid PET, dopamine transporter SPECT, and CSF proteomic studies, which currently are not readily available in Canada for clinical use. However, we anticipate that these biomarkers will be added to the current standard of care as they are further validated. Future reviews will address these novel imaging and proteomic technologies. Th ese advances will allow for a more aggressive therapeutic and personalized medicine approach that may impact on the management of the more common EODs.