Alzheimer Disease: New Concepts on Its Neurobiology and the Clinical Role Imaging Will Play

Learning Objectives:

After reading the article and taking the test, the reader will be able to:

• • List the relationships among pathologic features, biomarkers, and clinical expression of Alzheimer disease

• • Describe the characteristic signature of Alzheimer disease on images obtained with different modalities

• • Explain how imaging biomarkers are used in Alzheimer disease clinical trials

Accreditation and Designation Statement

The RSNA is accredited by the Accreditation Council for Continuing Medical Education (ACCME) to provide continuing medical education for physicians. The RSNA designates this journal-based activity for a maximum of 1.0 AMA PRA Category 1 Credit™. Physicans should claim only the credit commensurate with the extent of their participation in the activity.

Disclosure Statement

The ACCME requires that the RSNA, as an accredited provider of CME, obtain signed disclosure statements from the authors, editors, and reviewers for this activity. For this journal-based CME activity, author disclosures are listed at the end of this article.

Introduction

The clinical manifestation of Alzheimer disease (AD) and its associated brain patholophysiology, senile plaques and neurofibrillary tangles, was described more than 100 years ago (1). For roughly the next three-fourths of a century, the disease was to some extent ignored; it was believed to be an uncommon condition and thus not of great public health concern. This view has changed substantially in recent years, and AD is now recognized as a potentially unprecedented public health problem.

This is driven by three forces: (a) the inexorable aging of the population in all industrialized societies, (b) the fact that the incidence of AD increases exponentially with age in people older than 60 years, and (c) the current absence of any disease-modifying interventions.

Demographics and Public Health Impact

AD is a progressive neurodegenerative condition that primarily affects cognition. AD is not a normal part of aging; however, old age is its single greatest risk factor. Dementia is defined as disease-related loss of memory and other cognitive abilities of sufficient severity to interfere with activities of daily living. While there are a number of possible causes of dementia, the most common cause by far in elderly persons is AD (2). The current estimated prevalence in the United States is 5.3 million individuals, 5.1 million of whom are over the age of 65 years (2). Approximately one-eighth (13%) of all individuals in United States over the age of 65 years and more than half of all persons older than 85 carry a clinical diagnosis of AD dementia. The incidence doubles every 5 years after the age of 60. The prevalence of AD is greater in women than men; however, this is due to the greater life expectancy of women. Because of population aging, the estimated prevalence in the United States alone will be 7.7 million by 2030 and 14 million by 2050. AD is the fifth leading cause of death for those aged 65 years and older (3).

In addition to the personal toll on patients and families, the disease has substantial economic consequences. AD is the most common illness leading to nursing home placement (2). Total payments for health care and nursing home expenses for 2010 were $172 billion, including $123 billion for Medicare and Medicaid. By the year 2050, the estimated cost of AD in the United States will exceed $1 trillion per year. The cumulative cost from 2011 through 2050 is estimated at $20 trillion (2). The public health problem is not limited to the United States. Estimates of global costs of the disease in 2009 were $604 billion per year, or 1% of the world’s gross domestic product. About 70% of people with dementia live at home. Most of these patients receive unpaid help from family members and friends. In 2009 in the United States alone, in addition to formally counted health care costs described above 10.9 million family and other unpaid caregivers provided an estimated 12.5 billion hours of care, valued at $144 billion (2).

Clinical Course of the Disease

Longitudinal studies performed over the past 20 years have demonstrated that in individuals who eventually develop AD dementia, the cognitive symptoms begin insidiously and, in most cases, progress gradually. A large literature on mild cognitive impairment (MCI) has arisen since the mid-1990s that documents the gradual impairments of cognitive function (4). MCI refers to the transitional phase between normality and clinically evident dementia (5). The critical clinical distinction between the MCI and dementia phase of AD is loss of functional independence as the defining feature of dementia. Other clinical rating scales likewise reflect gradual development of clinical symptoms (6). The gradual onset and prolonged time course of AD stands in sharp contrast to many other medical conditions that have a clearly definable clinical onset, and often resolution.

Longitudinal studies show that a clinical diagnosis of MCI is a significant risk factor for the subsequent development of AD dementia. The rate of conversion from MCI to AD dementia is approximately 12% per year, which stands in contrast to the 1%–2% per year rate for cognitively normal elderly persons (7). Within 5 years, approximately 80% of individuals diagnosed with MCI will have converted to dementia (5).

Pathologic Features and Their Relationship to Clinical Symptoms

Two abnormal protein aggregates characterize AD pathologically. These proteinopathies are the basis of the well-known amyloid plaques and neurofibrillary tangles (Fig 1). Neurofibrillary tangles are intracellular aggregates, while amyloid plaques form in the extracellular space. Other important pathologic features are periplaque inflammatory reaction and neurodegeneration, which is characterized by synapse loss, neuron loss, and gross cerebral atrophy (25–28).

The neurofibrillary pathologic process of AD begins in the transentorhinal area and progresses to the hippocampus, then to paralimbic and adjacent medial-basal temporal cortex, to neocortical association areas, and lastly to primary sensorimotor and visual areas (29). Tau protein is a component of the cytoskeletal microtubule system. In AD, tau becomes hyperphosphorylated, disassociates from microtubules, assumes a paired helical filament configuration, and forms insoluble neurofibrillary tangles inside neurons (30).

The second major proteinopathy associated with AD is Aβ deposition. The hallmark Aβ peptide deposit in AD is the neuritic plaque, which consists of a dense central Aβ core with inflammatory cells and dystrophic neurites in its periphery. The core is made up of Aβ, which has aggregated into β pleated sheets (31). Amyloid precursor protein is a normal transmembrane protein. Normal proteolytic metabolism of amyloid precursor protein by secretases results in several peptide fragments (32). The most important of these are the Aβ_1–40 and Aβ_1–42 fragments (33). Of the two, Aβ_1–40 is more abundant, while Aβ_1–42 is more prone to aggregate (34). Soluble oligomeric Aβ fibrils are generally thought to be the neurotoxic Aβ species.

The field has seen intense debate on the question of which of the two hallmark proteinopathies is causative, Aβ or tau. Clinical examination–autopsy correlation studies (30,35–37) have demonstrated a much tighter correlation between neurofibrillary pathologic processes and cognitive impairment than between amyloid pathologic processes and cognitive impairment. Conversely, the available genetic evidence cited above strongly implicates a derangement in Aβ metabolism as the primary instigating factor in AD. Autopsy findings in cognitively normal subjects are informative. Approximately 30% of cognitively normal elderly subjects have AD pathologic features at autopsy (38–40). In most cases, cognitively normal subjects who meet autopsy criteria for AD have neocortical diffuse amyloid plaques but not extensive neocortical neurofibrillary tangles. These observations have been interpreted to support the position that amyloid develops first, while neocortical neurofibrillary tangles appear later (37).

AD Biomarkers

A biomarker is a physiologic, biochemical, or anatomic parameter that can be objectively measured as an indicator of normal biologic processes, pathologic processes, or responses to a therapeutic intervention. The term is used in the AD field to describe both biofluid analytes and imaging measures. Biofluid analytes in this context can refer to proteins in any biofluid; however, cerebrospinal fluid (CSF) biomarkers are presently the most well developed. At present, there are five AD biomarkers that have been widely studied and are well enough validated to be commonly considered for use in clinical trials. Two of these are CSF analytes and three are brain imaging measures (Table). These five biomarkers fall into two major mechanistic categories: (a) biomarkers of Aβ accumulation, which are represented by abnormal radiotracer retention on amyloid PET images and low levels of CSF Aβ₄₂, and (b) biomarkers of neuronal degeneration or injury, which are indicated by elevated levels of CSF tau (both total and phosphorylated); decreased FDG uptake on PET images in the temporoparietal cortex; and atrophy on structural MR images in the medial, basal, and lateral temporal lobes and the medial and lateral parietal cortices. Other potential AD biomarkers have been summarized in recent reviews (41,42); some of these are measurements from imaging methods that will be discussed later as potential future applications.

Both CSF Aβ₄₂ and amyloid PET findings are biomarkers of brain Aβ plaque deposition (Fig 2). Nearly all subjects who have received a clinical diagnosis of AD have positive amyloid imaging studies (43–45). Excellent correspondence has been seen between antemortem amyloid PET imaging findings and Aβ deposition in the brain (or cerebral vasculature) at autopsy (46,47). A consistent finding across different studies is that around 30% of cognitively normal elderly subjects have abnormal findings from PET amyloid imaging studies, which matches quite well the proportion of cognitively normal elderly subjects with an autopsy diagnosis of AD, as described above (48–52). Older age, APOE*E4 genotype, and a family history of AD are all associated with greater amyloid PET binding in cognitively normal late-middle-aged to elderly subjects (22,53,54). Measurable increases in amyloid load over time are seen in some cognitively normal elderly subjects (55–57). Although most amyloid imaging investigations published to date have used carbon 11 Pittsburgh compound B (PiB) as the tracer, newer fluorine 18 (¹⁸F)-labeled compounds (58–61) will have a higher clinical effect owing to the greater availability of ¹⁸F-labeled ligands. Depressed levels of CSF Aβ₄₂ correlate with both the clinical diagnosis of AD and Aβ pathophysiology at autopsy (62–64). Extremely high concordance is present between an abnormally low level of CSF Aβ₄₂ and positive PiB PET amyloid imaging findings in subjects who have undergone both tests (50,65–69). This is taken as evidence that these two biomarkers are, to a substantial extent, equivalent measures of brain Aβ pathophysiology.

CSF tau is an indicator of tau pathophysiology and associated neuronal injury. While phosphorylated tau may be a more specific indicator of AD, both phosphorylated tau and total tau levels increase in AD (70). Elevated CSF tau can be seen in other conditions, such as stroke and head trauma, and therefore is not specific for AD, but it does map to clinical disease severity (71,72). In AD, the cause of elevated tau level in the CSF is speculated to be a direct result of a tau pathophysiology that disrupts neurons, particularly axons, thereby releasing cytoskeletal elements, including tau, into the extracellular space, which appear in the CSF (73,74). Elevated CSF tau correlates with neurofibrillary pathophysiology at autopsy (75).

FDG PET measures of glucose metabolism reflect net brain metabolism, which predominantly reflects synaptic activity (76,77). In the context of AD, decreased FDG uptake is an indicator of impaired synaptic function. FDG PET studies in AD subjects demonstrate a specific topographic pattern of decreased glucose uptake in a lateral temporoparietal and a posterior cingulate-precuneus distribution (78,79) (Fig 3). Imaging-autopsy correlation studies have demonstrated good correlation between the antemortem FDG PET diagnosis of AD and postmortem confirmation (80). Clinically asymptomatic middle-aged carriers of the APOE*E4 gene, who are at increased risk for developing AD dementia later in life, have reduced FDG uptake in an AD-like pattern (81–83).

The structural MR imaging findings in AD are a measure of cerebral atrophy, which reflects microscopic neurodegeneration (loss of synapses, dendritic processes, and neurons) (84). Volumetric or voxel-based measures of brain atrophy display tight correlation between the severity of atrophy and the severity of cognitive impairment (72,85–88). Rates of brain atrophy correlate with rates of concurrent cognitive decline (89,90) (Fig 4). Atrophy on MR images is not specific for AD, but the degree of hippocampal atrophy correlates well with Braak staging at autopsy (29,91–93). The topographic distribution of atrophy on MR images (medial, basal, and lateral temporal lobe and medial parietal cortex) maps well with Braak staging of neurofibrillary tangles in subjects who have undergone antemortem MR imaging and postmortem AD staging (94,95).

Temporal Ordering of AD Biomarkers

On the basis of the observations that AD biomarkers provide information about different AD-related pathophysiologic processes and the idea that these processes may be temporally ordered, great interest has arisen recently in the idea of staging disease by using biomarkers. Several biomarker-based models of disease have been proposed (57,96–98). One biomarker-based disease model (57) was based on empiric observations from serial MR imaging studies and Pittsburgh compound B PET studies (Fig 5). This model (57) embodies the following data-derived (51,57) principles: (a) The presence of brain amyloidosis is necessary but not sufficient to produce cognitive decline; (b) the neurodegenerative component of AD, rather than the amyloid component, is the direct substrate of cognitive impairment, and the rate of cognitive decline is driven by the rate of neurodegeneration; and (c) the rates of amyloid deposition and of neurodegeneration are dissociated: Amyloid deposition is an early event, and neurodegeneration is a later occurring event. These principles are supported by biomarker data (51,57), as well as autopsy data, which indicate that the aspect of the AD pathologic process that is most closely coupled with cognitive impairment is neurodegeneration, particularly synapse loss (26,40,99). In this amyloid and neurodegeneration model (57), which was based on longitudinal and cross-sectional MR imaging and amyloid PET imaging data, clinical symptoms do not appear until fairly late in the disease process (Fig 5).

An expanded version of this disease biomarker model (96), illustrated in Figure 6, incorporates all five of the most well-validated AD biomarkers into a hypothetical comprehensive sequence of biomarker events as subjects progress from cognitively normal in middle age to dementia in old age. This hypothetical model rests on the assumption that these five AD biomarkers become dynamic in a sequential manner. The hypothesis is that amyloid biomarkers—PET amyloid imaging findings and CSF Aβ₄₂ measures—depart markedly from normality 15 or more years before the first clinical symptoms appear. Biomarkers of neuronal injury—CSF tau and FDG PET findings—become dynamic later. Structural MR imaging is the most dynamic of the five major biomarkers in the clinically symptomatic phase of the disease, where the MR findings correlate best with clinical symptom severity. Preliminary empiric evidence supports this concept of temporal ordering of AD biomarkers that lies at the heart of the model (100–104). This model follows the principle that none of the biomarkers are static; rates of change in each biomarker vary over time and follow a nonlinear time course, which is hypothesized to be sigmoid shaped. The hypothetical sigmoid-shaped trajectory was based in part on the observation that many measurements have floor and ceiling effects with a roughly linear response in the midregion. It was also based in part on biologic considerations. A sigmoid shape as a function of time implies that the biomarker experiences an initial acceleration phase and a later deceleration phase, with the inflection midpoint in the sigmoid curve representing the onset of deceleration (105).

Modifiers of Clinical Expression: Cognitive Reserve and Genetics

A feature of the biomarker model presented in Figure 6 is the lag between Aβ plaque formation and the neurodegenerative cascade followed by clinical symptoms. Intersubject variation in these lag periods is likely due to a number of factors. One factor is person-to-person differences in Aβ processing or in the effects of soluble oligomeric Aβ on neuronal injury. Many of these are likely under genetic control. Recent genome-wide association studies have identified new susceptibility loci for late onset AD (other than APOE) that implicate pathways related to brain cholesterol metabolism and inflammation (106–109).

A major modifier of the clinical expression of underlying AD pathophysiology is brain resiliency, or cognitive reserve (110). Cognitive reserve is an inclusive term that encompasses different mechanisms, including the numbers of neurons and synapses, the sensitivity of neurons and glia to pathologic processes, neuroplasticity, and the ability of a person to engage alternative cognitive processing strategies in the face of pathologic brain insult (110). Reserve can be expressed as the difference between the predicted and the observed cognitive performance of an individual for a given level of brain pathophysiology (111). The principle that subjects with higher reserve have a greater capacity to cope with pathologic insults than do those with low reserve is supported by data from several sources. Epidemiologic studies consistently show that higher educational and occupational attainment is protective against dementia (112–116). Clinical-autopsy studies have shown that at a given level of AD pathologic findings at autopsy, highly educated individuals are less likely to manifest clinical symptoms of dementia than are less-educated individuals (117,118). Similar findings have been reported in vivo, where among individuals with elevated Pittsburgh compound B uptake on PET amyloid imaging studies, cognitive performance was better with increasing education (119–121). Authors of a recent report (122) also suggest that lifetime cognitive enrichment may reduce amyloid deposition in later life.

Modifiers of Clinical Expression: Coexistent Age-related Brain Abnormality

A core concept essential to understanding relationships between brain pathophysiology and cognitive impairment in elderly persons is the frequency of comorbid brain pathologic conditions in this age group. Community-based clinical-autopsy studies have consistently demonstrated that a substantial majority of elderly subjects who received a diagnosis of AD dementia in life have pathologic conditions in addition to AD at autopsy. The most common of these are cerebrovascular disease (CVD) and Lewy body disease (123,124). The reason mixed conditions are so common is straightforward. As is the case with AD, processes such as Lewy body disease and CVD increase in prevalence with age. The older an individual is, the more likely he or she is to have accumulated different independent brain pathologic conditions.

While pure vascular dementia is uncommon, vascular contributions to cognitive impairment are very common (125,126). CVD is generally regarded to be the second most common pathologic contributor, after AD, to dementia in elderly persons (123,127,128). The most common vascular lesion found in community autopsy studies is microinfarction (124,127,129,130). Evidence to date indicates that the effects of AD abnormalities and vascular brain injury on cognitive expression are additive (131–133). Vascular risk factors likewise increase the risk of developing dementia (134–137). This has led to the recommendation to engage in a “heart healthy” lifestyle as a means of avoiding or delaying the onset of dementia.

The defining neuropathologic feature of dementia with Lewy bodies is pathologic aggregation of α-synuclein protein in neurites to form Lewy bodies (138). Lewy body disease is the proteinopathy underlying both Parkinson disease and dementia with Lewy bodies (139). Lewy body disease is commonly found as a comorbid condition, along with AD and CVD, in demented subjects examined in community autopsy studies (127,130). Hippocampal sclerosis, grain disease, and TDP-43 proteinopathy are also common coexistent conditions at autopsy, although not as common as AD, CVD, and Lewy body disease (140).

One condition that may seem to merit greater coverage in a review on dementia is frontotemporal lobar dementia (FTLD) (141). FTLD describes a family of neurodegenerative disorders characterized by focal lobar degeneration of the frontal and/or temporal lobes. FTLD is actually as common as AD in subjects under the age of 60 years (142,143); however, in individuals older than 65, AD and Lewy body disease account for the majority of neurodegenerative dementias, while FTLD is far less common.

Diffusion, Perfusion, Spectroscopy, and Functional MR Imaging

It might seem odd to find discussion of diffusion, perfusion, spectroscopy, and functional MR imaging toward the end of a review on AD biomarkers. However, although these MR imaging techniques have received substantial attention from investigators, none is well enough established to be included as a biomarker in newly formulated diagnostic criteria for AD (144–149), nor are any of these techniques routinely used in AD clinical trials. These techniques might best be regarded presently as of academic interest and awaiting sufficient validation to be accorded wider clinical use. Each of these techniques will be briefly described below.

Applications of Imaging Biomarkers: New Diagnostic Criteria for AD

On the basis of the preceding material, one might assume that CSF and imaging AD biomarkers currently play a prominent role in daily clinical practice. It might be surprising, however, to realize that this is not the case. Several explanations are typically offered for this state of affairs. Perhaps the most common is the fact that the absence of approved disease-modifying therapies minimizes the impetus for early diagnosis because early diagnosis will not change clinical management. An obvious reason that biomarkers are not widely used in clinical practice is that until very recently, diagnostic guidelines for AD did not incorporate CSF or imaging biomarkers. Criteria for the clinical diagnosis of AD were originally established in 1984 by a work group comprising representatives from the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (234). In the 1984 criteria, AD dementia was a diagnosis made solely on clinical grounds. It was also operationalized as a diagnosis of exclusion: AD was diagnosed when other potential causes of dementia were excluded. The role of imaging in the 1984 criteria (234) was, therefore, to help exclude conditions like cerebral infarction, space-occupying lesions, or potentially treatable causes such as hydrocephalus or subdural hemorrhage.

Recently, however, diagnostic guidelines for AD have been updated (145–149). One of these efforts was organized by the National Institute on Aging–Alzheimer’s Association. In 2010 these organizations commissioned three work groups. Each group was assigned the task of defining or revising diagnostic criteria for one of three recognized phases of the disease: preclinical AD, symptomatic predementia or MCI, and AD dementia (146–149). Biomarkers are used in all three phases of the disease. In addition to their traditional exclusionary role, imaging studies are now expected to provide positive diagnostic information. In the preclinical phase biomarkers are by definition the only indicator of latent disease. They are used to establish the presence of AD pathophysiologic processes in subjects with no overt symptoms. In the symptomatic MCI and AD phases, biomarkers are used to establish the underlying cause for an observed clinical deficit. In addition to aiding in establishing a pathophysiologic etiology to aid diagnosis, imaging and CSF biomarkers provide two additional major classes of clinically useful information. These are prediction of future clinical course (235–243) and longitudinal measurement of disease progression (244).

Applications of Imaging Biomarkers: Clinical Trials

In addition to use as diagnostic criteria in the new diagnostic guidelines (145–149), a second major use category for imaging and CSF biomarkers is in therapeutic trials. Incorporation of imaging and CSF biomarkers into clinical trial design is more advanced than is their incorporation in clinical diagnosis. For several years, nearly every therapeutic trial in AD has incorporated one or more AD biomarkers in the study design, where they can serve one or more specific functions.

Need for Standardization of Imaging

The clinical practice of radiology is based on visual assessment of images. This diagnostic assessment is converted to a narrative interpretation which becomes the radiologic contribution to a patient’s diagnostic evaluation. Visual assessment of the degree of hippocampal atrophy on MR images (266) (Fig 4) or the degree of temporoparietal hypometabolism (Fig 3) is easily implemented and widely available. However, anatomic or physiologic processes are continuous phenomena, and verbal encoding of a visual impression does not lend itself to accurate or reproducible assessment of fine incremental grades of atrophy or glucose uptake.

For years, academic efforts have focused on quantitative measurements—that is, turning gray-scale images into numbers—for statistical analysis in research papers. This is particularly true for MR imaging. Initial efforts at image quantification focused on methods based on regions of interest, which were often hand drawn. In MR imaging, these efforts often focused on specific named structures such as the hippocampus or entorhinal cortex, which are involved early and progressively in AD (267,268). More recently, methods have been developed to automatically parcellate (269) gray matter density or the thickness of cortical surfaces (270–272) into regions of interest. This approach is reproducible and does not require manual intervention. Several investigators have developed multivariate analysis and machine learning–based algorithms that use the entire three-dimensional MR imaging data set to form a disease model against which individual subjects may be compared. A new incoming study is scored according to the degree and topographic pattern of atrophy as compared with the images from a large database of well-characterized subjects (273–279). Such measures capture the severity of neuronal abnormality—that is, Braak staging over the whole brain is better than measurement of hippocampal volume for determining severity (95).

Of the five major biomarkers of AD, three are imaging studies. Although the value of imaging biomarkers in both clinical diagnosis and clinical trials is clear, major barriers exist to actual implementation. The most important barrier has been lack of standardized methods, particularly methods for extracting quantitative information from images. Although initiatives such as the Alzheimer’s Disease Neuroimaging Initiative have focused on standardization of image acquisition methods (280,281), universally accepted standards for image quantification have not yet emerged.

An important contribution that the imaging community could make would be to establish open-source methods for standardized quantification. If imaging is to take its place as a widely used, clinically relevant disease biomarker for AD, then imaging should transform itself to operate as do biomarkers in other fields. For example, measures of blood pressure, serum glucose, or serum lipids have been standardized. Experts in the fields of hypertension, diabetes, and lipid metabolism have agreed on and adopted diagnostic classification criteria that are based on standardized laboratory measurements. Treatment protocols are tied to these universal measures. For imaging to assume a similar role as an AD biomarker, similar standardization of quantitative measures should take place. In an ideal world, a subject could be imaged with a unit by any vendor. An analysis software pipeline would be used to analyze that subject’s scan, extract quantitative data, compare those data to a standardized database, and assign the subject a probabilistic differential diagnosis and an imaging-based disease severity rank (282). Ideally this would happen “online” on the imaging unit. These quantitative diagnostic imaging measures would be incorporated into universally recognized diagnostic criteria for AD.

Disclosures of Potential Conflicts of Interest: Financial activities related to the present article: none to disclose. Financial activities not related to the present article: Institution has received money from Pfizer, Johnson & Johnson, Janssen, Eisai, and Elan for consultancies. Other relationships: none to disclose.