Seminar Vol 397 April 24, 2021 1577

Alzheimer’s disease Philip Scheltens, Bart De Strooper, Miia Kivipelto, Henne Holstege, Gael Chételat, Charlotte E Teunissen, Jeffrey Cummings, Wiesje M van der Flier

In this Seminar, we highlight the main developments in the field of Alzheimer’s disease. The most recent data indicate that, by 2050, the prevalence of dementia will double in Europe and triple worldwide, and that estimate is 3 times higher when based on a biological (rather than clinical) definition of Alzheimer’s disease. The earliest phase of Alzheimer’s disease (cellular phase) happens in parallel with accumulating amyloid β, inducing the spread of tau pathology. The risk of Alzheimer’s disease is 60–80% dependent on heritable factors, with more than 40 Alzheimer’s disease-associated genetic risk loci already identified, of which the APOE alleles have the strongest association with the disease. Novel biomarkers include PET scans and plasma assays for amyloid β and phosphorylated tau, which show great promise for clinical and research use. Multidomain lifestyle-based prevention trials suggest cognitive benefits in participants with increased risk of dementia. Lifestyle factors do not directly affect Alzheimer’s disease pathology, but can still contribute to a positive outcome in individuals with Alzheimer’s disease. Promising pharmacological treatments are poised at advanced stages of clinical trials and include anti-amyloid β, anti-tau, and anti-inflammatory strategies.

Introduction Alzheimer’s disease is the main cause of dementia and is quickly becoming one of the most expensive, lethal, and burdening diseases of this century.1 Since the Seminar published in 2016,2 important developments have taken place in the under standing of the underlying pathology, the recognition of multiple causative and protective genes, the identification of new blood-based and imaging biomarkers, and the first cautious signals of positive effects of disease-modifying treatments and lifestyle interventions. The aim of this new Seminar is to provide the reader with up to date insight into the field of Alzheimer’s disease.

Clinical signs and symptoms Three cases, in panel 1 (see also figure 1), illustrate the clinical spectrum of Alzheimer’s disease. Case A highlights Alzheimer’s disease that is determined genetically, as per the ongoing global initiatives of the Dominantly Inherited Alzheimer Network and Alzheimer Prevention Initiative and their associated clinical trials. Case B represents a language variant of Alzheimer’s disease, usually occurring at a younger age (under 70 years), illustrating the difficulty in recognising Alzheimer’s disease in those for whom memory problems are not the first and most prominent feature. Case C is a typical amnestic variant, more com- monly seen in patients older than 70 years, illustrating the growing population affected by Alzheimer’s disease and dementia: older individuals often living alone, and increasingly dependent on others for care.

Diagnostic criteria: from clinical, to clinical and biological, to biological The diagnosis of Alzheimer’s disease has gone from a purely pathological one, in the days of Alois Alzheimer (1864–1915), to a clinical, exclusionary approach in 1984. The clinical diagnosis was based on the criteria defined by the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association,3 via a combined clinical and biological approach developed by the International

Working Group4,5 and subsequent efforts by the National Institute on Aging and the Alzheimer’s Association working groups,6 incorporating biomarkers to make the categorisation of Alzheimer’s disease purely biological.7 Initially, the diagnosis of Alzheimer’s disease was restricted to the stage of dementia, a clinical syndrome characterised by substantial progressive cognitive impair- ment affecting several domains, or neurobehavioral symptoms of enough severity to cause evident functional impact on daily life. A person with dementia is no longer fully independent, and this loss of independence is the primary feature differentiating dementia from mild cognitive impairment.8

Given the developments in the biomarker field and the desire to make them usable in a diagnostic setting, Jack and colleagues8 grouped the biomarkers into A (amyloid), T (phosphorylated tau), and N (neurodegeneration, mea sured by total tau where applicable): the ATN frame work (appendix p 1). In this research framework, the diagnosis of Alzheimer’s disease is defined by the presence of amyloid β and phosphorylated tau. The

Lancet 2021; 397: 1577–90

Published Online March 2, 2021 S0140-6736(20)32205-4

Alzheimer Centre Amsterdam (Prof P Scheltens MD, H Holstege PhD, Prof W M van der Flier PhD), Department of Neurology (Prof P Scheltens), Department of Clinical Genetics (H Holstege), Department of Clinical Chemistry (Prof C E Teunissen PhD) and Department of Epidemiology and Datascience (Prof W M van der Flier), Amsterdam University Medical Centers, Amsterdam, Netherlands; Life Science Partners, Amsterdam, Netherlands (Prof P Scheltens); VIB Center for Brain and Disease Research, Leuven, Belgium (Prof B De Strooper MD); KU Leuven Department for Neurology, Leuven, Belgium (Prof B De Strooper); Dementia Research Institute, University College London, London, UK (Prof B De Strooper); Division of Clinical Geriatrics and Center for Alzheimer Research, Department of Neurobiology, Care Sciences and Society, Karolinska University Hospital, Stockholm, Sweden (Prof M Kivipelto MD); Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finland (Prof M Kivipelto); Ageing and Epidemiology Research Unit, School of Public Health, Imperial College London, London, UK (Prof M Kivipelto);

Normandie Université, Université de Caen, Institut National de la Santé et de la Recherche Médicale, Groupement d’Intérêt Public Cyceron, Caen, France (Prof G Chételat MD); Chambers-Grundy Center for Transformative Neuroscience, Department of Brain Health, University of Nevada, Las Vegas, NV, USA (Prof J Cummings MD); and Cleveland Clinic Lou Ruvo Center for Brain Health, Las Vegas, NV, USA (Prof J Cummings)

Search strategy and selection criteria

Between Dec 1, 2019, and Sept 1, 2020, we searched the Cochrane Library for articles published exclusively in English during 2010–15, PubMed for articles published during 2016–20, and Embase for articles published during 2016–20. We used the search term “Alzheimer’s disease” in combination with the following: “pathology”, “imaging”, “diagnosis”, “therapy”, “trials”, “epidemiology”, “CSF”, “genetics”, and “biomarkers”. We largely selected publications from the past 5 years, and especially focused on changes that occurred after the publication of the previous Seminar in 2016.1 We also searched the reference lists of articles identified by this search strategy and selected those that were judged relevant. Review articles and book chapters are cited to provide readers with references for more details than this Seminar can include.




1578 Vol 397 April 24, 2021

Correspondence to: Prof Philip Scheltens, Department

of Neurology, Amsterdam University Medical Center,

Amsterdam 108HX, Netherlands

See Online for appendix

presence of amyloid β (regardless of the presence of phosphorylated tau and neurodegeneration) is termed Alzheimer’s pathological change, basing the research diagnosis of Alzheimer’s disease on biomarker evidence only. Clinical stages can range from cognitively normal to mild cognitive impairment and dementia, stressing the continuum of Alzheimer’s disease, which spans a period of years. The ATN framework underpins the importance of amyloid β and tau as the defining characteristics of Alzheimer’s disease, conse quently proposing that Alzheimer’s disease can be diagnosed by biomarkers only, and definitively distin guishing between

the concepts of Alzheimer’s disease and dementia (figure 2).

Despite the critique that other key causes of dementia, in particular vascular disease, were omitted,9 the authors of the ATN framework argued that dementia has multiple underlying pathologies, of which Alzheimer’s disease is one, but Alzheimer’s disease is defined by the presence of amyloid β and tau (acknowledging that many other pathologies can also be present in these patients).10 The large number of ATN categories, combined with the fact that other pathologies are not evaluated in the scheme, makes the ATN approach not

Panel 1: Case vignettes

Mrs A, aged 42 years, a successful manager of an IT company, presents at the Alzheimer Centre Amsterdam because of self-perceived memory loss and loss of oversight and multitasking abilities. She recognises these complaints all too well because her mother had Alzheimer’s disease for 5 years, until her death at the age of 47 years. Two of her four brothers also had Alzheimer’s disease, and had been tested and found to be carriers of a PSEN1 mutation. Although she has not been tested herself, she always felt she would be a carrier and subsequently chose not to have children. She asked for a full evaluation because she wanted to have the option of participating in a clinical trial programme. Her Mini-Mental State Exam score was 27/30 and her Montreal Cognitive Assessment score was 24/30. Given her age, these scores suggest mild memory and executive disturbances, which were confirmed by neuropsychological testing. A brain MRI showed no abnormalities. Cerebrospinal fluid biomarker values were 750 pg/mL for amyloid β42, 335 pg/mL for tau, and 35 pg/mL for phosphorylated tau 181, all in the abnormal range. Serum neurofilament light chain value was 25 pg/mL, which is abnormal for her age, according to in-house defined reference curves. APOE status was ε3/ε4. All these biomarker values indicate the presence of Alzheimer’s disease pathology and onset in a clinically mildly affected patient. Genetic testing confirmed the presence of the same PSEN1 mutation carried by her brothers. She was informed about the diagnosis, followed up at 6 month intervals at the centre, and put on the list for a clinical trial within the Dominantly Inherited Alzheimer Network Trials Unit programme. She informed her colleagues at work and agreed to have regular meetings with the company’s physician.

Mr B, aged 62 years, is a high school teacher who presented to the neurologist with gradually progressive difficulty finding words and understanding sentences, and slight memory loss. He had visited another neurologist because of suspicion of a vascular event, but a brain MRI showed no abnormalities. On examination, his Mini-Mental State Exam score was 25/30 and the Montreal Cognitive Assessment score was 24/30, both within normal range for his age, with normal findings at routine neurological and laboratory investigations. Neuropsychological and detailed language assessment revealed a decrease in fluency, naming, and repetition of long sentences.

Review of the MRI showed slight asymmetry of the temporal lobes, with grade 2 hippocampal atrophy on the left side and grade 1 hippocampal atrophy on the right side, without any other abnormalities (figure 1B). Because of his young age, and his and his family’s desire to obtain a firm diagnosis to plan ahead and make proper adjustments to his working life, an amyloid-PET scan was done and showed diffuse cortical uptake of the ligand (figure 1A). As part of a research project, a tau-PET scan was done and showed left-temporal abnormal tau deposition (figure 1C). A diagnosis of logopenic variant of Alzheimer’s disease was made. Lifestyle advice was given and regular visits to a speech therapist were offered.2 Given the diagnosis and the perceived grim future, as well as the high demands of his job on his language skills, he decided to take sick leave from his job.

Mrs C, aged 78 years, lives independently on her own after being widowed 6 years ago. She was known to her general practitioner with controlled hypertension and moderate heart failure, for which she takes medication. Her son lives abroad and her daughter lives 100 km away. Both have demanding jobs and young children. During telephone and Skype calls, her children noticed increasing forgetfulness and one of the neighbours had recently informed the daughter that her mother mixed up the days, forgot to eat, and was not able to take good care of herself anymore. The daughter accompanied her mother to the Alzheimer Centre Amsterdam on referral by the general practitioner, who had initially dismissed the worries of the daughter. On examination by a geriatrician, she was found to be malnourished and underweight. The Mini-Mental State Exam score was 17/30 and a brief neuropsychological test battery showed scores below the norm for memory and executive function. Her score on the Amsterdam Instrumental Activities of Daily Living test3 was 58, indicating severe impairment. An MRI showed a medial temporal atrophy score of 2 bilaterally, and moderate to severe white matter changes (Fazekas score 2). A diagnosis of mild to moderate dementia due to Alzheimer’s disease with some vascular contribution was made, and a case manager was assigned to organise and supervise care to have her stay at home as long as possible. Vascular risk factors were checked and cholinesterase inhibitor therapy was started.



Seminar Vol 397 April 24, 2021 1579

yet suitable for clinical practice.11 In addition, there are operational limitations to defining A, T, and N positivity or negativity, such as some biomarkers not having established cutoff points, and different biomarkers being combined in one category. Although the AT or ATN approach is the cornerstone of current trials of disease-modifying interventions in Alzheimer’s disease, clinical diagnosis still rests on the criteria set by the National Institute on Aging in 2011.6,12

The ATN framework clearly paves the way for a diagnosis before the stage of Alzheimer’s disease-associated demen- tia, and it makes individualised risk-profiling for patients with mild cognitive impairment feasible.13 However, a clinical encounter study evaluating doctor–patient com- munication in memory clinics showed that clinicians are reluctant to share specific prognostic information with patients with mild cognitive impairment.14 In the context of predementia diagnosis, subjective cognitive decline is even more challenging. A recent Personal View provides a clinical characterisation of subjective cognitive decline, and attempts to provide clinicians with guidance on how to deal with this decline (which might or might not be attributable to underlying Alzheimer’s disease).15 At a group level, ATN biomarkers clearly predict incident dementia in subjective cognitive decline, but individualised risk modelling remains challenging.16,17 In a Delphi study to identify topics most relevant to discuss in the diagnostic process, patients and caregivers indicated that they value precise and specific information, even when it does not provide complete certainty.18 Tools to support decision making and communication about Alzheimer’s disease diagnosis, such as ADappt,19 are urgently needed.

Epidemiology Incidence and prevalence In 2018, Alzheimer’s Disease International estimated a dementia prevalence of about 50 million people worldwide, projected to triple in 2050, with two-thirds living in low-income and middle-income countries.20 The most recent data estimate that dementia prevalence in Europe will double by 2050.1 Accumulating evidence suggests that the incidence of dementia is declining in high-income countries,21 although evidence for a decline in prevalence is less convincing.22

Mortality The relatively stable prevalence despite decreasing incidence could be explained by a long disease duration, although studies on mortality do not support this notion. A US-based study evaluating survival after a dementia diagnosis in almost 60 000 individuals reported survival times of 3–4 years.23 In an European, memory clinic- based cohort, median survival time was 6 years after a diagnosis of Alzheimer’s disease dementia (median 6·2 years [range 6·0–6·5]).24 This estimate coincides with a multicentre study that provided estimates of duration not only of the dementia stage, but also of the prodromal (mild cognitive impairment) and of preclinical disease stage of Alzheimer’s disease.25 For an individual aged 70 years, duration estimates are 10 years for the preclinical stage, 4 years for the prodromal stage, and 6 years for the dementia stage of Alzheimer’s disease, totalling 20 years. A first attempt at estimating prevalence on the basis of a biological (rather than clinical) definition showed that, at the age of 85 years, the prevalence of biologically defined Alzheimer’s disease is 3 times higher than that of clinically defined Alzheimer’s disease.26

Risk factors for dementia and Alzheimer’s disease The strongest risk factors for Alzheimer’s disease are advanced age (older than 65 years, although this is not a fixed definition) and carrying at least one APOE ε4 allele.27 Moreover, women are more likely to develop Alzheimer’s disease than are men, especially after the age of 80 years.20

Figure 1: Imaging findings of a case similar to patient B’s case in panel 1 (A) Amyloid Pittsburgh compound B-PET scan showing amyloid deposition predominantly in the posterior cingulate region. (B) T1-weighted MRI images showing generalised cortical atrophy, left to right. (C) Tau-PET image using AV1451 tracer, showing left-sided inferotemporal lobe, parietal, and mild posterior cingulate deposition of tau. Image courtesy of Rik Ossenkoppele and Gil Rabinovici.


For ADappt see www.adappt. health

Alzheimer’s disease

Normal DementiaMild cognitive impairment

15–25 years

Figure 2: Alzheimer’s disease is a continuum The arrow points to the continuum of Alzheimer’s disease, stretching over a period of 15–25 years, in which Alzheimer’s disease pathology can be present without any symptoms via a stage of mild cognitive impairment leading up to overt dementia, illustrating that dementia is the end result of a long-time presence of Alzheimer’s disease pathology. Not every patient will necessarily follow this path by definition. Note: between normal and mild cognitive impairment, patients can experience subjective complaints, but not all complaints are early signs of dementia and the predictive value of having complaints for dementia is unknown.




1580 Vol 397 April 24, 2021

Women are also more likely to have a higher tau load, despite having a similar amyloid β burden.28,29 In addition, cardiovascular risk factors and an unhealthy lifestyle have been associated with an increased risk of dementia. The Lancet Commission on Dementia Prevention esti- mated that 12 modifiable risk factors together account for roughly 40% of the worldwide risk of any type of dementia.30 These estimates illustrate that prevention by intervening on modifiable risk factors is of great relevance, even if most of the dementia burden cannot be prevented via a lifestyle-intervention approach. However, evidence suggests that vascular risk factors do not increase the risk of Alzheimer’s disease pathology as measured by cerebrospinal fluid biomarkers or PET.31–33 This evidence implies that lifestyle and vascular risk factors contribute to dementia, but not via the Alzheimer’s disease pathway.

Genetics Causative and risk genes Studies of twins showed that the risk of Alzheimer’s disease is 60–80% dependent on heritable factors.34 The common APOE ε4 allele explains a substantial part of, but does not completely account for the heritability of, Alzheimer’s disease.35,36 Large genome-wide association studies have been done to identify novel genetic variants in Alzheimer’s disease, the latest of which to date investigated about 150 000 people with Alzheimer’s disease and age-matched controls, and more than 300 000 people with a proxy-phenotype Alzheimer’s disease (parental history of Alzheimer’s disease) and controls (no parental history of Alzheimer’s disease), which increased the number of Alzheimer’s disease- associated risk alleles to more than 40.37 However, although the common APOE ε4 risk allele is associated with an estimated 3–4 times increased risk of Alzheimer’s disease across different genome-wide association studies, other Alzheimer’s disease risk alleles are associated with much smaller contributions to the total disease risk (odds ratio between 1·05 and 1·2; figure 3B).37

Based on the presence or absence of these risk alleles in the genome of an individual, a polygenic risk score can be calculated, which is currently able to distinguish between patients with Alzheimer’s disease and controls with 75–85% accuracy.38,39 Although the bulk of this accuracy can be ascribed to the APOE ε4 allele, the 40 or so other variants also collectively contribute substantially to Alzheimer’s disease risk.27 Functional annotation of these risk loci indicate that, next to amyloid β metabolism, the modulation of the immune response, cholesterol, lipid dysfunction, endocytosis, and vascular factors play a role in the development of Alzheimer’s disease.40–45 Next-generation sequencing techniques have shown rare protein-damaging variants in the SORL1,46 ABCA7,47 and TREM2 genes.48,49 These findings suggest that the intact protein products of these genes are essential in maintaining brain health (figure 3A).

Protective genes The identification of risk-increasing genetic variants has fuelled the interest in the detection of protective genetic variants (figure 3C). Carriers of the protective APOE ε2 allele have an estimated 2 times decreased lifetime risk of Alzheimer’s disease compared with non- carriers,50 which translates into an exceptionally low likelihood of Alzheimer’s disease for homozygous APOE ε2 allele carriers.51 The discovery of the rare Ala673Thr Icelandic protective mutation of APP52 was associated with prolonged cognitive health. Similarly, compared with middle-aged individuals, a rare Pro522Arg amino acid change in the PLCG2 gene was associated with a near 2 times reduced risk of Alzheimer’s disease53 and other types of dementia, and with a 2·3 times increased chance of reaching 100 years in cognitive health.54,55 Genetic resilience was even reported in a person with a PSEN1 mutation who lived beyond the age of onset of symptoms common in her family, due to a homozygous rare protective variant in the APOE ε3 allele (Christchurch mutation).56 Variants in the klotho longevity gene were associated with a similar effect.57 Such protective genetic variants hold great promise in Alzheimer’s disease research, as they might pinpoint mechanistic processes protecting cognitive health.

Pathophysiology Basic scientists designate the preclinical phase of Alzheimer’s disease as the cellular phase. Alterations in neurons, microglia, and astroglia drive the insidious progression of the disease before cognitive impairment is observed.58 Neuro-inflammation,59 alterations in the vessels,60,61 ageing,62 and dysfunction of the glymphatic system63 act upstream or in parallel to accumulating amyloid β in this cellular disease landscape. Amyloid β induces, via an unknown way, the spread of tau patho logy,64 which is associated with the appearance of necroptosis markers in neurons displaying granulo- vacuolar degeneration.65

Single-cell transcriptome analysis has elucidated the microglia response.66 APOE and TREM2, two major Alzheimer’s disease risk genes, are important parts of this response.66–68 ApoE binds to amyloid β plaques,69 and the Alzheimer’s disease-associated genetic variants of TREM2 Arg47His, Arg62His, and Asp87Asn decrease binding of TREM2 to ApoE (figure 3).70 Several other proteins linked to genetic risk of Alzheimer’s disease, such as SHIP1, CD2AP, RIN3, BIN1, PLCG2, CASS4, and PTKB2 act presumably downstream of ApoE and TREM2 signal- modulating endocytosis, motility, and phagocytosis in microglia (figure 4). CD33 acts in opposition to TREM2,77 and MS4A4A modulates the secretion of soluble TREM2 protein.78 The fact that so many Alzheimer’s disease risk genes converge on microglial response pathways indicates their central role in the disease pathogenesis. However, further research is needed to elucidate whether the microglia response is to amyloid β plaques only,76 or



Seminar Vol 397 April 24, 2021 1581

that it also mediates toxicity induced by tau pathology79 or acts protectively against tau.80

The contradictory effects of the microglia response partly reflect the limitations of mice models overexpressing tau for the study of Alzheimer’s disease. It is possible that strong transgenic tau overexpression79 induces an artificially strong neuroinflammatory response that is not seen in milder tau models.76,80 The use of mice models that do not overexpress tau,76 mouse-human chimeric mice,82,83 or new in-vitro models derived from human, induced pluripotent stem cells84 might help to explain the conflicting observations. Of note, all preclinical models are reductionistic in nature, implying that any conclusions towards therapeutic developments need to be made with caution.

Although cellular pathology has become central in the study of Alzheimer’s disease, great progress has also been made in understanding the preceding biochemical phase

of the disease (in ATN terms, before A positivity [presence of amyloid β]). Thanks to cryo-electron microscopy, amyloid β85 and tau fibrils are now known in finer detail.86 Cryo-electron microscopy has also allowed full insight into how presenilins, the catalytic subunits of γ-secretases, interact with APP87 and Notch substrates.88 Complemented by func tional studies on purified γ-secretase complexes,71 it is now understood that clinical mutations in presenilins destabilise the γ-secretase–APP interactions, leading to premature release of longer, aggregation-prone amyloid β peptides. These insights support the development of new therapeutic approaches to tackle amyloid β in Alzheimer’s disease.

The role of amyloid β in the disease cascade needs to be reintegrated with concepts of resilience and susceptibility. To this end, the cellular responses of neurons, astroglia, microglia, pericytes, and endothelial cells, which are largely defined by the genetic makeup of a patient, will

Figure 3: The genetic landscape of Alzheimer’s disease MAF (x-axis) is the frequency at which a non-reference (variant) allele occurs in the population. Variant carriers with OR=1 and non-carriers have the same odds of developing Alzheimer’s disease, variants with OR >1 are associated with an increased risk of Alzheimer’s disease, and variants with OR <1 are associated with a protective effect (y-axis). (A) Causative or strong risk increasing variants. A schematic representation of individual rare variants for which ORs cannot be estimated due to extreme variant rareness. Linkage studies in large pedigrees indicate that specific rare variants in PSEN1, PSEN2, and APP cause autosomal dominant Alzheimer’s disease, in some cases with age at onsets as early as 40 years old. Note that not all variants in these three genes give rise to autosomal dominant Alzheimer’s disease; some might be risk-modifiers or non-pathogenic. Further, evidence is accumulating that certain variants in the SORL1 gene are causative of Alzheimer’s disease before the age of 70 years. The Alzheimer’s disease-association of variants in the SORL1, ABCA7, and TREM2 genes was found in gene-based tests; carriers may come from small pedigrees with inheritance patterns of Alzheimer’s disease suggestive of autosomal dominant inheritance. (B) GWAS hits are common (by convention, MAF >1%) variants that represent risk alleles that occur with significantly different frequency in patients with Alzheimer’s disease and controls. Each variant is represented …