This position paper by Dubois et al. is a long overdue call for reassessing the criteria, assumptions, and procedures for diagnosis of Alzheimer disease. The recommendations of this work group are well reasoned and sound. However, the task of crafting “consensus” criteria will not be easy for several reasons: a) lack of a clear definition of the “disease” or clinical phenomenon; b) tremendous heterogeneity in both clinical and biological phenotypes of the “disease”; and c) lack of or poorly understood causal relationship(s) between the clinical expression of the “disease” and their biological underpinnings.

In spite of these long-standing problems the field faces, the work group has done an excellent job in laying the foundation for an ongoing effort to refine the criteria that have been used since 1984.

View all comments by Zaven Khachaturian

The authors are to be congratulated on this paper! It is very timely, the call to incorporate a biological footprint in diagnosis is appropriate, and the elimination of MCI to AD as a binary outcome is most welcome. However, from my perspective, the salient information regarding dementia diagnosis is that an individual has declined in cognitive abilities, relative to previous levels, and that the decline is sufficient to interfere with function in everyday activities. Most importantly, then, we need to know how an individual has changed relative to his/her premorbid cognitive function—this is the principle of intraindividual change. The Core Diagnostic Criterion A proposed by Dubois et al. specifies that there be objective deficit on an episodic memory measure, presumably in comparison with age- and education-matched norms. Rather than judging that individuals have declined relative to previously attained abilities, their test performances are compared with the performances of other persons. Dependence of this interindividual comparison to determine the presence of dementia is...  Read more

The authors are to be congratulated on this paper! It is very timely, the call to incorporate a biological footprint in diagnosis is appropriate, and the elimination of MCI to AD as a binary outcome is most welcome. However, from my perspective, the salient information regarding dementia diagnosis is that an individual has declined in cognitive abilities, relative to previous levels, and that the decline is sufficient to interfere with function in everyday activities. Most importantly, then, we need to know how an individual has changed relative to his/her premorbid cognitive function—this is the principle of intraindividual change. The Core Diagnostic Criterion A proposed by Dubois et al. specifies that there be objective deficit on an episodic memory measure, presumably in comparison with age- and education-matched norms. Rather than judging that individuals have declined relative to previously attained abilities, their test performances are compared with the performances of other persons. Dependence of this interindividual comparison to determine the presence of dementia is flawed on several levels: 1) it does not indicate that the individual's performance has changed from his/her prior performance; 2) it does not indicate whether the cognitive test results relate to impaired everyday function; and 3) depending on where the cutpoint is established, in a normally distributed population some percentile of nondemented individuals will be classified as "impaired." For example, if the cutpoint is 1 SD of the mean, 17 percent will be below the tail; if the cutpoint is 2 SD of the mean, 2.5 percent will be below the tail. On the other side of the coin, some clearly dementing individuals of high intellectual attainment may not be classified as impaired based on cognitive test performance because they still are able to perform at a level above the cutpoint. Moreover, there are many potential confounds (e.g., education, literacy, culture) that affect interpretation of neuropsychological test performance.

I recognize that my views about the operationalization of diagnostic criteria for MCI/AD are very much in the minority, as the diagnosis most often is on cognitive test scores. I contend, however, that this test-based approach inherently identifies some individuals as having "MCI," for example, when in fact their cognitive abilities have not changed and they continue to function normally. Depending on where the cutpoints are set and the population to which they are applied, the proportion of such non-impaired individuals in the sample will vary, contributing directly to the reported variability in the outcomes of these samples. Basing the diagnosis of impairment on the principle of intraindividual change, on the other hand, is both highly accurate and sensitive in the early detection of symptomatic AD (Storandt et al., 2006). In fact, for the subset of individuals characterized with MCI for whom underlying AD is the cause, MCI is not a risk factor for AD; it is AD in its earliest symptomatic stage. Revised criteria for AD should take into account this knowledge and discard the concept of MCI in favor of etiological diagnoses; for most cases, MCI is early-stage AD (Morris, 2006 ).

View all comments by John Morris

This study by Richard et al. provides important findings on the diagnostic benefit of combining neuropsychological memory assessment with biomarkers for predicting AD dementia in patients with amnestic MCI.

The authors conclude that biomarkers are not useful to improve predictive accuracy once memory test performance has been accounted for. Since memory tests are less expensive than biomarker assessment, neuropsychological tests might be preferred in clinical practice.

The study has several strengths and supports the notion that the costs of biomarkers need to be weighed against their real benefit. It should be noted, however, that the current paper focused on a few measures, and none of the prediction models (biomarkers or memory tests) reached clinically relevant levels of accuracy (remaining below 80 percent). Thus, more sophisticated prediction models need to be developed, and biomarkers may still contribute to any improvement in the models.

Other points to consider are the predictive value with respect to differentiating AD from other types of dementia, which...  Read more

This study by Richard et al. provides important findings on the diagnostic benefit of combining neuropsychological memory assessment with biomarkers for predicting AD dementia in patients with amnestic MCI.

The authors conclude that biomarkers are not useful to improve predictive accuracy once memory test performance has been accounted for. Since memory tests are less expensive than biomarker assessment, neuropsychological tests might be preferred in clinical practice.

The study has several strengths and supports the notion that the costs of biomarkers need to be weighed against their real benefit. It should be noted, however, that the current paper focused on a few measures, and none of the prediction models (biomarkers or memory tests) reached clinically relevant levels of accuracy (remaining below 80 percent). Thus, more sophisticated prediction models need to be developed, and biomarkers may still contribute to any improvement in the models.

Other points to consider are the predictive value with respect to differentiating AD from other types of dementia, which was not assessed in the current study. Furthermore, it remains to be tested whether biomarkers are beneficial at the very early stage of AD, when no or only very slight cognitive impairment is present.

View all comments by Michael Ewers

The findings of this study are not that surprising—baseline cognition is the biggest predictor of cognitive progression in MCI. The RAVLT is the most sensitive test at the MCI stage simply because verbal recall is the key criterion used to define MCI patients for entry.

The Achilles' heel of most AD biomarkers is that their additive predictive value above and beyond a memory test is still not optimal. This is where we need to improve our tests. That said, pathology biomarkers have a high negative predictive value. MCI can be due to 20 different types of brain pathology, all yielding memory test scores in the same range; a negative brain scan can help clinicians rule out the likelihood of AD. In addition, memory tests are less useful at predicting future disease in people who have perfectly normal memory, whereas biomarkers can identify silent pathology. For example, in a recent multicenter study, amyloid PET significantly predicted an 18-month cognitive decline in normal subjects, above and beyond their baseline memory test score (see Doraiswamy et al., 2012).

The...  Read more

The findings of this study are not that surprising—baseline cognition is the biggest predictor of cognitive progression in MCI. The RAVLT is the most sensitive test at the MCI stage simply because verbal recall is the key criterion used to define MCI patients for entry.

The Achilles' heel of most AD biomarkers is that their additive predictive value above and beyond a memory test is still not optimal. This is where we need to improve our tests. That said, pathology biomarkers have a high negative predictive value. MCI can be due to 20 different types of brain pathology, all yielding memory test scores in the same range; a negative brain scan can help clinicians rule out the likelihood of AD. In addition, memory tests are less useful at predicting future disease in people who have perfectly normal memory, whereas biomarkers can identify silent pathology. For example, in a recent multicenter study, amyloid PET significantly predicted an 18-month cognitive decline in normal subjects, above and beyond their baseline memory test score (see Doraiswamy et al., 2012).

The overall message of the paper is a good one—that we need to develop biomarkers with significant additive value above and beyond simple memory tests.

References:
Doraiswamy PM, Sperling RA, Coleman RE, Johnson KA, Reiman EM, Davis MD, Grundman M, Sabbagh MN, Sadowsky CH, Fleisher AS, Carpenter A, Clark CM, Joshi AD, Mintun MA, Skovronsky DM, Pontecorvo MJ, AV45-A11 study group. Amyloid-β assessed by florbetapir F 18 PET and 18-month cognitive decline: a multicenter study. Neurology. 2012 Oct 16;79(16):1636-44. Abstract

View all comments by P. Murali Doraiswamy