The “Alzheimerization” of Dementia Research

In this issue of the Journal, Koopmans et al.1 present data on survival to “late-stage dementia” in a large cohort of Dutch nursing home patients. This study touches on important issues related to the clinical diagnosis of dementia and deserves careful scrutiny. Dementia case definitions are increasingly being criticized for their biases favoring the clinical features of Alzheimer's disease (AD), at the expense of non-AD conditions.2 This so-called “Alzheimerization” of dementia research is pervasive.3 For example, most dementia diagnostic systems require the presence of memory loss, making this domain “essential” to the diagnosis. However, there is only a weak association between AD pathology in either the hippocampus or the temporal cortex (both important memory structures) and measures of global cognitive impairment4 or clinicians' ratings of dementia.5, 6 The lesions that are most relevant to these outcomes must lie elsewhere in the brain. The American Academy of Neurology (AAN) recently published an evidence-based review of published criteria for the diagnosis of dementia.7 One striking finding was the conclusion that none of the most widely used dementia diagnostic systems achieved its highest evidence standard. In its call for future research, the AAN specifically felt that “memory should not be a required part of the diagnosis of dementia.” Nonetheless, dementia is screened for with memory tests and staged with instruments, such as the Clinical Dementia Rating (CDR) Scale8 that give special emphasis to the severity of memory complaints. This issue even applies to notions about what to expect as dementia “progresses.” Koopmans et al.'s concept of “late dementia” presumes that, regardless of its cause, dementia cases will progress toward a state of total disability if patients survive that long. Their data set, one of the few that follow institutionalized persons longitudinally, can provide information on this assumption. A unique and interesting feature of AD pathology is its hierarchically arranged progression in space (and presumably time) through the brain.9 AD pathology consistently begins in or near the hippocampus and only later moves out to affect the frontal, temporal, and parietal lobes. This sequence explains the consistent appearance of memory loss before other cognitive deficits in AD and the clinical features of its later stages. Moreover, to the extent that AD dominates clinical experience of dementia, the sequence is also consistent with expectations for the temporal progression of cognitive impairments, formally described by Reisberg et al.10 However, the fact that AD pathology alone has been shown to follow the sequence described by Braak et al.9 suggests that this progression is only relevant to AD. In other words, the clinical stages of dementia's progression, implied by Reisberg et al.'s stages and codified by the CDR, may have less validity outside of AD. Other patterns of progression, or even a lack of progression, may be possible, and this in turn may explain the heterogeneity in progression that is frequently observed in clinically diagnosed “AD” patients who are followed longitudinally. In this context, Koopmans et al.'s data take on new significance. They report that a minority of demented institutionalized persons ever reach the “final stages” of dementia, (a clinical state similar to stage 7 in Reisberg et al.'s system). Although clinically diagnosed AD cases dominate this group, all dementia types appear capable of reaching this stage. Unfortunately, this cannot be confirmed except by autopsy. Nonetheless, Koopmans et al.'s data offer several interesting observations. First, only a minority of demented persons appear destined for the “late dementia” that a knowledge of the Braak et al. hierarchy might predict. Thus, either most of these nursing home patients did not, in fact, have AD or clinically diagnosable AD progresses to its terminal stages only in a minority of cases. It cannot be easily assumed that Knoopmans et al. correctly classified their cases at baseline. The validity of AD's clinical diagnosis at autopsy has a surprisingly poor track record, especially in advanced old age. For example, the specificity at autopsy for a clinical diagnosis of National Institute of Neurological Communicative Disorders and Stroke—Alzheimer's Disease and Related Disorders Association “Probable AD” made after age 79 may be as low as 22%.11 It can be particularly difficult to clinically distinguish the dementia of AD from the dementia of frontal systems vascular disease when it is superimposed on trivial subclinical (but common) AD pathology.6, 12, 13 Braak et al. find that only a minority of even octogenarians have sufficiently advanced AD pathology (e.g., Braak Stage IV or higher) to explain the presence of dementia, independently of comorbid conditions (including normal aging) that affect cognition.9 The low conversion rate to “late dementia” in these cases could be trivially explained if most AD patients die of intercurrent illness before they reach this end-point. However, AD victims are often healthier than their peers, and Koopmans et al.'s data failed to show that the presence of comorbid vascular disease reduced the fraction of cases with AD who reach late dementia. In contrast, any non-AD diagnosis, be it vascular dementia or other conditions, appeared to protect patients from late dementia. Conversely, clinical features, such as increased baseline age and female sex, which are known risk factors for AD, increased the risk of late dementia. Future studies need to confirm their diagnoses against Braak's stages. It has been understood for many years that the mere presence of AD pathology does not imply dementia, but it is not clear what is a clinically significant AD burden. The transition to dementia may involve a discrete set of relatively few brain regions6 It is not known whether widely accepted pathological criteria for AD's diagnosis reflect a Braak stage in advance of or past the involvement of these critical regions. Many nondemented older persons have AD pathology.14 In fact, as many as half of “pathologically” diagnosed AD cases are not clinically demented.15 Braak staging offers the possibility of empirically determining the level of pathology required before the dementia of AD is observed. Moreover, it may be possible to achieve Braak staging in vivo.16, 17 This could greatly enhance the accuracy of clinical diagnoses and could be used to document the efficacy of future disease-modifying treatments.