ALZHEIMER'S DISEASE: ITS DIAGNOSIS AND PATHOGENESIS Jillian J. Kril I Centre for Education and Research on Ageing, Concord Hospital Department of Medicine, The University of Sydney, Concord, New South Wales, Australia 2139, and Department of Pathology, The University of Sydney, Sydney, New South Wales, Australia 2006 Glenda M. Halliday Prince of Wales Medical Research Institute Randwick, New South Wales, Australia 2031 I. Introduction II. Diagnostic Issues A. A/3 Plaques and NFTs for Pathological Diagnosis B. Evaluation of Other Pathologies C. Clinical Correlates of AD Pathology D. Reproducibility of Current Clinical Diagnostic Protocols E. Summary III. Pathogenesis A. Brain Atrophy B. Neuronal Loss C. A/3 Deposition D. NFT Formation E. Mechanisms of Degeneration E Summary IV. Genetic Influences A. Dominant Inheritance B. Genetic Risk Factors C. Summary V. Inflammation and Anti-inflammatory Drugs VI. Estrogen Therapy VII. Vascular Pathology in AD A. Vascular Risk Factors B. Summary References lAuthor to whom correspondence should be addressed. INTERNATIONAL REVIEW OF 167 Copyright © 2001 by Academic Press. NEUROBIOLOGY, VOL 48 All rights of reproduction in any form reserved. 0074-7742/01 $35.00 168 JILLIAN J. KRIL AND GLENDA M. HALLIDAY I. Intr(xludion In the century since the neuronal inclusions [neurofibrillary tangles (NFTs); Fig. 1] and extracellular protein aggregates (Aft plaques; Fig. 1) that form the pathological hallmarks of Alzheimer's disease (AD) were de- scribed, our knowledge of all aspects of AD has grown markedly. AD is uniformly progressive and ultimately results in debilitating cognitive im- pairment. In the early stages, the impairment may only be apparent on Lesion type "jAI~ plaques 0 e ; O 0 e ; • ; o • " D Lewy bodies . Clinical diagnosis m° ° t ? other dementi~ •''QI° l~c,. 1. The major pathologies resulting in dementia are neurofibrillary tangles (NFTs; top left), A¢I plaques (top center), and Lewy bodies (top right, arrow). Diagnosis of AD was pre- viously based on age-corrected densities of A~ plaques; however, the finding that a significant number of patients with dementia with Lewy bodies (DLB) also have A¢I plaques questions this practice. Similarly, NFTs can be found in other forms of dementia and thus are not specific for AD. Newer criteria proposed for the pathological diagnosis of AD use both A/~ plaques and NFTs. This change in the way in which AD is diagnosed pathologically will have a significant impact on the clinical criteria for the identification of AD. These clinical criteria have been val- idated using A~ plaque-based pathology and now require re-evaluation in light of the advances in our understanding of the pathogenesis of the disease. ALZHEIMER'S DISEASE 169 neuropsychological testing; however, by end stage, few functions above the automatic remain unaffected (Forstl and Kurz, 1999). The pattern and se- quence of functional deficits and the relentlessness of the decline are rel- atively reproducible, with the course of the disease in one patient similar to that in others. AD remains the most prevalent form of late-life dementia and is the most significant cause of morbidity in the elderly. Yet, we are still unable to, in most instances, accurately predict who will succumb to AD or to effectively treat it in those who do. II. Diagnostic Issues The definitive diagnosis of AD is made by pathological examination of the brain; however, the accurate and reproducible diagnosis of AD during life is of paramount importance. Not only is it essential to exclude possible treatable causes of dementia, but it is also necessary for the identification of homogeneous groups of patients for evaluation and study, and for the recruitment of patients for drug and other therapeutic trials. Variability in the clinical diagnosis of AD is well recognized and major diagnostic issues continue to be addressed to develop accurate and reproducible criteria for its identification. Yet, similar issues for the neuropathological diagnosis of AD are only beginning to be evaluated in a systematic fashion. In most in- stances, neuropathology is considered the "gold standard" for the diagnosis of AD, but considerable variation exists between diagnostic protocols. This has the potential to have a significant impact on our understanding of the disease. A. A/3 PLAQUES AND NFTs FOR PATHOLOGICAL DIAGNOSIS The majority of protocols for the pathological diagnosis of AD use only one of the major pathological lesions first described by Alzheimer. The most commonly used lesion for the diagnosis of AD is the neuritic plaque as Aft deposition is more distinctive for AD than other neurodegenerative diseases. The Consortium to Establish a Registry for Alzheimer's Disease (CERAD; Mirra et al., 1991) developed the most widely used diagnostic pro- tocol. It employs the earlier National Institutes of Health protocol based on age-corrected plaque densities (Khachaturian, 1985) in a semiquantitative fashion to arrive at diagnoses of varying certainties. The CERAD criteria (Mirra et al., 1991; Table I) has gained wide acceptance because of its ac- curacy and simplicity. In 142 cases with a clinical diagnosis of probable AD 170 JILLIAN J. KRIL AND GLENDA M. HALLIDAY TABLE I CONSORTIUM TO ESTABLISH A REGISTRY FOR ALZHE1MER'S DISEASE (CERAD) CRITERIA FOR NEUROPATHOLOGICAL DIAGNOSIS OF AD a Plaque Densities b <50 Years 50-75 Years > 75 Years N + No dementia N + No dementia N + No dementia, or S + no dementia S, M, or F + No dementia S + Dementia Not defined M + Dementia S, M, or F + Dementia Normal brain Possible AD S + No dementia Probable AD S + Dementia, or M + no dementia Definite AD S, M, or F + Dementia F + Dementia aRegions examined middle frontal, superior and middle temporal, inferior parietal cor- tices, hippocampus and entorhinal cortex, midbrain. bN = none, S = sparse, M = moderate, F = frequent plaques. From Mirra et aL (1991). (the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS)-Alzheimer's Disease and Related Disorders Association (ADRDA) most certain clinical category, see Section II.C and Table II), 84% were found to have definite AD 7% probable and 2% possible (Mirra et al., 1991). In 7% of cases, the age-related plaque score was negligible, suggesting another cause for dementia in these cases. In an independent assessment of the accuracy of the CERAD criteria, subjects with a clinical diagnosis of prob- able AD had definite AD at autopsy in 27 out of 28 cases (96% Kosunen et al., 1996). Subsequent interlaboratory testing of the CERAD criteria revealed 75% agreement in the rank ordering of 10 cases between 24 neuropathol- ogists at 18 centers (Mirra et al., 1994). The majority of variation was due to staining differences between laboratories, but the data suggest there is considerable variation between pathologists in the CERAD diagnosis of in- dividual cases. In contrast to the plaque-based protocols, Braak and Braak (1991) pro- posed a staging scheme for the neuritic pathology ofAD (NFTs and neuropil threads). This six-stage scale documents the temporal sequence and topo- graphic spread of AD pathology. Although not proposed as a criteria for the neuropathological diagnosis of AD, dementia is reliably associated with stages V and VI and to a variable degree with stages III and IV (Braak and Braak, 1991; Harding et al., 2000). NFTs first form in the pre-a layer of the transentorhinal cortex (stage 1, Table III), then the pre-a layer of the entorhinal cortex, the hippocampus, and finally the isocortex (Table III). Compared with the CERAD criteria, inter-rater reliability for neuritic stag- ing is high (weighted kappa 0.85 to 0.97; Nagy et al., 1997b). However, the ALZHEIMER'S DISEASE 171 TABLE II NINCDS-ADRDA CRITERIA FOR CLINICAL DIAGNOSIS OF An a Possible AD Probable AD Key features Probable AD Supportive features Probable AD Suggestive features Probable AD Features not consistent with AD Definite AD Clinical diagnosis of • Dementia syndrome with variation in onset, presentation, or course but in absence of neurological, psychiatric, or systemic disorder sufficient to cause dementia • Dementia in presence of second systemic or brain disorder sufficient to produce dementia, but not considered to be cause of dementia • Single, severe, progressive deficit in single cognitive domain Inclusion Criteria • Onset between 40 and 90 years • Dementia established on clinical examination and testing • Deficits in two or more areas of cognition • Progressive worsening of functions Exclusion Criteria • No disturbance of consciousness • Absence of systemic or brain disease, which may account for progressive deficits in memory and cognition • Progressive deterioration in specific cognitive function, such as language, motor skills, and perception • Impaired activities of daily living and altered behavior pattern • Family history of similar disorder • Normal CSF • Normal or nonspecific changes on EEG • Progressive cerebral atrophy on CT scan After exclusion of other causes • Plateaus in course of progression • Associated symptoms of depression, insomnia, incontinence, delusions, illusions, hallucinations, catastrophic verbal, emotional or physical outbursts, sexual disorders, weight loss • Neurological abnormalities, especially in advanced patients motor signs (increased muscle tone, myoclonus, or gait disorder), epilepsy • CT scan noixnal for age • Sudden apoplectic onset • Focal neurological signs hemiparesis, sensory loss, visual field deficit, incoordination early in disease • Seizures or gait disturbance at onset or early in disease • Clinical criteria for probable AD • Histopathologic evidence from biopsy or autopsy aDiagnostic certainty is ranked as possible, probable, or definite based on the features present. From McKhann et al. (1984). 172 JILLIAN J. KRIL AND GLENDA M. HALLIDAY TABLE III BRAAK STAGING SCHEME OF NFT FORMATION DURING AGING AND AD Stage Neurofibrillary tangles a I II III 1V V VI NFTs in pre-a layer of the transentorhinal cortex Isolated NFTs in pre-a layer of entorhinal cortex proper Numerous NFTs in transentorhinal cortex Sparse NFTs in CA1 sector of hippocampal formation Severe involvement of transentorhinal cortex, including eNFTs Modest involvement of CA1 NFTs in subiculum Numerous NFTs in CA1, some in CA4 Mild involvement of isocortices, sparing of primary cortices All sectors of hippocampus involved Moderate involvement of subcortical nuclei Isocortex moderately involved Severe involvement of isocortices Severe involvement of subcortical nuclei Mild involvement of primary cortices aNFT = neurofibrillary tangle, eNFT = extracellular or "ghost" neurofib- rillary tangle. From Braak and Braak (1991 ). strict hierarchical order of NFT formation is not observed in all cases. In a study of 42 brains, Gertz and colleagues (1998) found that only six cases fully fitted the expected pattern of NFT distribution. Most of these violations of staging order were in the early stages, suggesting they would not alter the effectiveness of the protocol for identifying AD. The NINCDS and ADRDA working group proposed a number of criteria for the neuropathological diagnosis ofAD, which evaluated both Aft plaques and NFT and excluded cerebrovascular disease (Tierney et al., 1988). How- ever, the protocols have not been widely adopted, in part, because of their complexity and modest sensitivity. Comparison between NINCDS-ADRDA clinical and pathological criteria showed agreement in 8 of 9 nondemented controls, 18 of 38 cases with possible AD, and 18 of 19 cases with prob- able AD (Nagy et al., 1998). An evaluation of the Khachaturian, CERAD, NINCDS-ADRDA, and Braak methods for assessing AD in a group of 60 el- derly subjects with known Mini-Mental State (MMS) score, revealed that all criteria accurately identified individuals with severe dementia (MMS 0-10; Jellinger et al., 1995). However, in moderately demented individuals (MMS 11-23) reliance on plaque-based criteria resulted in significant underdiag- nosis of AD. The majority of these subjects also had limbic NFTs (Braak stages III and IV). In addition, the use of plaque densities without assessing clinical state resulted in 5 of 9 nondemented subjects being classified as AD, ALZHEIMER'S DISEASE 173 indicating the presence of plaques is a poor indicator of the presence of dementia, at least in the very old. Because of these difficulties, the neuritic staging scheme of Braak has been combined with the CERAD protocol for assessment of plaques into the National Institute on Aging (NIA)-Reagan Institute criteria for the diagnosis of AD (Hyman and Trojanowski, 1997; National Institute on Aging and Reagan Institute Working Group, 1997; Newell et al., 1999). Topographical assessment of NFT type (intra- or extracellular) and A/3 plaque density is used to classify cases into high, intermediate, or low likelihood of AD. These criteria are the currently accepted "gold standard" for the diagnosis of AD, even though their reliability, reproducibility, and overall accuracy are yet to be determined. B. EVALUATION OF OTHER PATHOLOGIES Further problems with assessing the accuracy of neuropathological di- agnoses exist. In most studies, diagnostic validity is assessed by reporting cases that meet criteria for AD, regardless of whether other pathologies are present. In some instances, coexisting pathologies may contribute to the de- mentia process and thus be of importance in the assessment of the clinical criteria. In one study addressing this issue, diagnostic accuracy was 81% for AD (using CERAD criteria) including coexisting disease, but only 44% for pure cases (Bowler et al., 1998). The presence of infarction was the primary reason for the differences encountered. Furthermore, the identification of a number of previously unrecognized dementing disorders, with clini- cal and/or pathological overlap with AD (e.g., dementia with Lewy bodies (DLB) ; Kosaka et al., 1984; McKeith et al., 1996, 1999; Fig. 1; and small vessel disease dementia (Pantoni et al., 1996)), calls into question the usefulness of many of the existing criteria for the diagnosis of AD. Because the majority of cases with DLB also have plaques (McKeith et al., 1996), the CERAD cri- teria cannot differentiate these disorders and the evaluation of intracellular pathology is required. In addition, the overlap between cerebrovascular disease and AD is well known (see Di Iorio et al., 1999; Breteler, 2000; de la Torre, 2000). However, the demonstration that small vessel disease alone can cause a clinical syndrome indistinguishable from AD (Pantoni et al., 1996) suggests that reevaluation of the role of this pathology is also necessary. Few studies have addressed these issues. The NIA-Reagan Institute criteria for the diagnosis of AD (Hyman and Trojanowski, 1997; National Institute on Aging and Reagan Institute Working Group, 1997) states that all pathologies should be evaluated, but does not suggest how overlapping diagnoses can be arrived at or how they 174 JILLIAN J. KRIL AND GLENDA M. HALLIDAY should be incorporated into the diagnostic criteria. This represents a signif- icant weakness in the current neuropathological criteria for the diagnosis of AD and must be addressed before the effectiveness of any criteria can be adequately evaluated. C. CLINICAL CORRELATES OF AD PATHOLOGY Reevaluation of the clinical diagnosis of AD in light of the changing con- cepts of the pathology of the disease is necessary. It is no longer adequate that clinical criteria for AD alone are developed. Better diagnostic crite- ria for patients with similar core clinical features that differentiate the signs and symptoms of other pathologies are required. The most widely used clin- ical criteria are those established by the NINCDS-ADRDA (McKhann et al., 1984). Diagnoses of probable and possible AD (Table II) can be made using these criteria. Possible AD is considered when dementia is apparent in the presence of other systemic or brain disorders that, may themselves, result in dementia. A diagnosis of probable AD is considered when the patient is free from complicating diseases and when deficits in two or more areas of cogni- tion are present. The CERAD group (Morris et al., 1989) proposed a battery of clinical and neuropsychological tests to aid in the classification of cases into the NINCDS-ADRDA possible and probable AD groups. A multicenter study of the NINCDS-ADRDA criteria, which evaluated 60 cases (40 with AD), showed an initial sensitivity of 0.81 and specificity of 0.73 (Blacker et al., 1994). These values were improved to 0.83 and 0.84, respectively, after consensus rating (Blacker et al., 1994). Validity studies that tested the NINCDS-ADRDA criteria against patho- logically confirmed cases show variable results depending on the cases in- cluded in the study. In "typical" cases, the agreement between probable and definite AD has been shown to be 100% (Martin et al., 1987; Morris et al., 1988). However, in unselected cases, accuracies of 68-76% and 88% for probable AD (Burns et al., 1990; Risse et al., 1990) and 78% for possible AD (Burns et al., 1990) were found. Thus, the NINCDS-ADRDA protocol for the diagnosis of AD has been well-validated within and across centers as correlating with the CERAD plaque-based pathological criteria. However, because these studies would have included cases with DLB and possibly other pathologies, reevaluation of the accuracy of these criteria is required. In addition to the NINCDS-ADRDA criteria for the clinical diagnosis of AD, a number of other diagnostic protocols for use in clinical and popu- lation settings have been developed. The Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) of the American Psychiatric Association (APA; World Health Organization, 1992) criteria require a ALZHEIMER'S DISEASE 175 deficit in memory, as well as one other cognitive domain of gradual onset and progressive decline. These criteria result in similar classification of pa- tients to the NINCDS-ADRDA criteria (see above). In addition, the Clinical Dementia Rating (CDR; McCulla et al., 1989; Morris, 1993) and Mini-Mental State Examination (MMSE; Folstein et al., 1975) are used to determine the severity of dementia. These protocols are useful for screening for cognitive impairment as they are easy to administer and have been validated across a variety of social and ethnic populations. However, they lack specificity for AD. Pathological validation of these protocols has been performed, but as with other criteria, reevaluation is necessary in light of the changing na- ture of dementia diagnosis. It would be of interest for those centers with large clinical and pathological databases to reevaluate the clinical diagnosis of AD and similar dementia syndromes using the currently recommended neuropathological criteria for AD. This may help define better diagnostic tools that can then be evaluated longitudinally. D. REPRODUCIBILITY OF CURRENT CLINICAL DIAGNOSTIC PROTOCOLS Several studies to test the reliability of the clinical criteria for AD have been performed. Both Lopez and colleagues (1990) and Kukall and col- leagues (1990) used four raters to evaluate the NINCDS-ADRDA criteria. Using cases with dementia (AD and non-AD) and nondemented controls, percentage agreement ranged from 55% to 75% for pairs of raters (kappa coefficients were 0.36-0.65) with the most experienced clinicians achiev- ing the greatest agreement (Lopez et al., 1990). Interestingly, many of the disagreements in diagnosis were found between possible and probable AD categories. Although there is little disagreement on whether patients have dementia, the underlying causes of the dementia syndrome are less reli- ably agreed upon between clinicians. The inclusion of cases with multiple pathologies using the current criteria probably contributes to this variability. E. SUMMARY Considerably more research is required on the diagnosis and definition of AD. The current recommended "gold standard" has yet to be widely validated, and protocols for overlapping pathologies need to be incorpor- ated. This will, of course, have an impact on the clinical diagnosis of AD. At present, clinical criteria for AD cannot differentiate patients with differ- ent underlying disease mechanisms (e.g., DLB versus AD). In addition, it will be important to determine the clinical profile of cases that have both 176 JILLIAN J. KRIL AND GLENDA M. HALLIDAY NFTs and Aft. Difficulties with the definition of the disease have important implications for any study of AD. From the point of view of researching the pathogenesis of AD, pure groups are desired to eliminate potential con- founding causes. The continual modification and improvement of criteria for the diagnosis of AD is therefore necessary until we understand, and either prevent or cure, this illness. IIh Pathogenesis To understand the pathogenesis of AD, it is necessary to determine the sequence of events that occurs over the life of a patient. Because it is not possible to perform longitudinal cellular analyses in humans, most of our understanding of the pathogenesis of AD is inferred from patients sampled cross-sectionally at different time points in the disease process. Information concerning the initial events is the most patchy as it is extremely difficult to determine, with accuracy, when the disease first begins. The single greatest risk factor for the development of AD is age (Jorm, 1990). However, we do not yet fully understand the normal aging process; thus, it is difficult to distin- guish the pathological process (es) underlying AD. We do know that demen- tia in old age isfar from a universal phenomenon and that other factors must play a role in determining susceptibility to disease. These factors include ge- netic, environmental, and lifestyle factors, as well as coexisting disease. Although a decline in brain function with age is accepted as normal by many authors, increasing evidence suggests cognitive decline is not an in- evitable consequence of aging (Rubin et al., 1998; Morris, 1999; Unger et al., 1999) but rather a manifestation of underlying disease processes. Longitudi- nal studies of community-dwelling elderly subjects do not find a decrease in cognitive performance with advancing age (Rubin et al., 1998; Morris, 1999). Interestingly, those who do develop dementia may have a long preclinical period with stable deficits (usually memory), which precedes a precipitous decrease in function (Rubin et al., 1998; Small et al., 2000). Such studies call into question the idea of an age-related decline in brain function and more likely represent cohort differences in health, education, and other factors. Nevertheless, many of these studies are performed on highly selected groups of elderly subjects who are free from neurological and systemic diseases, and although adequately addressing the question of age-associated cognitive de- cline, do not represent the majority of elderly subjects. Cognitive deficits may be present in a proportion of elderly subjects, although these would be expected to have greater brain pathology. Numerous studies have shown an increased risk of AD in subjects with low education levels (primary school level or around 6 or less years of schooling) . performed. Both Lopez and colleagues (19 90) and Kukall and col- leagues (19 90) used four raters to evaluate the NINCDS-ADRDA criteria. Using cases with dementia (AD and non-AD) and nondemented controls,. (Mirra et al., 19 91; Table I) has gained wide acceptance because of its ac- curacy and simplicity. In 14 2 cases with a clinical diagnosis of probable AD 17 0 JILLIAN J. KRIL AND GLENDA M. HALLIDAY. stages V and VI and to a variable degree with stages III and IV (Braak and Braak, 19 91; Harding et al., 2000). NFTs first form in the pre-a layer of the transentorhinal cortex (stage 1, Table