Revised Criteria for Mild Cognitive Impairment May Compromise the Diagnosis of Alzheimer Disease DementiaRevised MCI Criteria Impact on AD Dementia ONLINE FIRST

Mild cognitive impairment (MCI) aims to identify cognitive decline in its earliest stages. It is a popular concept: the National Institute on Aging in 2004 noted MCI as a top discovery of the Alzheimer's Disease Centers (ADCs) program,¹ and in 2010, more than 1220 publications listed MCI in the title, abstract, or key words (Elsevier SciVerse Scopus). Given its importance, the National Institute on Aging and the Alzheimer's Association convened a work group to update criteria for MCI. The revised criteria recently were published,² as were revised criteria for Alzheimer disease (AD) dementia³ and for the preclinical stage of AD.⁴ A key criterion for MCI is the absence of dementia.² The revised criteria for AD dementia, which are comparable with widely used criteria for dementia of the Alzheimer type,⁵ stipulate that impairment must be present in 2 or more cognitive domains and must interfere with the ability to function in usual activities. The original diagnosis of MCI^{6 - 7} was limited to individuals with cognitive impairment in a single domain (memory), thus distinguishing MCI from dementia, but more recently its differentiation from dementia has come to rest solely on the preservation of functional activities.^{8 - 9} The revised criteria for MCI,² however, allow considerable latitude as to what represents functional independence and thus blur the categorical distinction between MCI and dementia.

Initially proposed as a predictor of progressive cognitive decline in older adults,¹⁰ MCI later was defined as an amnestic syndrome that denoted a transitional stage between aging and dementia (particularly dementia caused by AD).^{6 - 7} With experience, however, came the realization that individuals classified as having MCI frequently were impaired in cognitive domains other than memory.^{11 - 12} Criteria for MCI thus were broadened to include both nonamnestic presentations and the involvement of multiple cognitive domains but continued to require “essentially normal” functional activities.^{8 - 9} The recently published revised criteria for MCI require (1) change in cognition recognized by the affected individual or observers; (2) objective impairment in 1 or more cognitive domains; (3) independence in functional activities; and (4) absence of dementia.² Although on the surface these revised criteria conform to earlier definitions,^{8 - 9} the revised criteria operationalize “independence in functional activities” more expansively than before. For example, “mild problems” in performing daily activities such as shopping, paying bills, and cooking are permissible, as is dependency on aids or assistance to complete such activities.² This interpretation of “independence in functional activities” thus has the potential to characterize some individuals who now are diagnosed with very mild and mild AD dementia as having MCI.

To assess this potential, the functional ratings of participants enrolled at federally funded ADCs with clinical and cognitive data maintained by the National Alzheimer's Coordinating Center (NACC) (grant U01 AG016976; W. Kukull, PhD, principal investigator) were evaluated. The degree of functional impairment as assessed by 2 ratings of activities of daily living, the Functional Activities Questionnaire (FAQ)¹³ and the Clinical Dementia Rating (CDR),¹⁴ was determined for NACC participants with MCI and for participants with very mild and mild AD dementia (all diagnoses were made prior to the publication of the revised criteria for MCI). These functional ratings then were interpreted in accordance with the definition in the revised criteria for MCI of “independence in functional activities” to examine their potential to change current diagnoses.

Data were obtained from participants entered into the NACC between September 2005 and May 20, 2011, who were diagnosed by ADC clinicians as being cognitively normal, having MCI, or having probable AD; standard criteria were used to diagnose MCI^{8 - 9} and probable AD.¹⁵ Diagnoses were made by a single clinician or by consensus conference. A uniform assessment battery to characterize individuals with MCI and probable AD in comparison with cognitively normal aging was developed by a Clinical Task Force of the ADCs and implemented at each ADC in September 2005. This Uniform Data Set (UDS)^{16 - 17} defines the standard clinical and cognitive observations that are collected longitudinally on ADC participants; the UDS data are submitted to the NACC. Incorporated into the UDS is the FAQ, a measure of the participant's functional impairment as rated by an informant. The FAQ rates the participant's current ability to perform 10 instrumental activities of daily living: (1) household finances; (2) assembling financial and tax documents; (3) shopping; (4) hobbies; (5) simple kitchen functions, such as heating water; (6) preparing a meal; (7) monitoring current events; (8) comprehending televised programs or a book; (9) monitoring family occasions or personal medications; and (10) transportation (eg, driving). Performance in each of the 10 activities is rated along a 4-point scale, from no difficulty in performance (0), to difficulty but still performing “by self” (1), to performing with assistance (2), to dependent (3). Only those activities that the participant previously performed are rated and, in the UDS, reflect impairment due only to cognitive loss. A conservative operationalization of the functional “independence” permitted by the revised criteria for MCI is considered herein to include FAQ scores of 0 or 1 for each rated activity. Because the revised criteria for MCI allow activities to be performed with assistance, also considered herein is a more liberal interpretation where FAQ ratings of 0, 1, and 2 could be compatible with this definition of functional “independence.”

The UDS also includes the CDR. Using both informant report and an examination of the participant, an experienced clinician rates the presence or absence of cognitive and functional loss and, when present, rates severity in each of 6 domains: Memory, Orientation, and Judgment and Problem Solving and function in Community Affairs (CA), Home and Hobbies (HH), and Personal Care. The clinician synthesizes all relevant information, including that obtained with the FAQ, in rating instrumental activities of daily living in the community (CA) and at home (HH). Impairment caused by cognitive loss only is rated on a 5-point scale as none (0), very mild (0.5), mild (1), moderate (2), or severe (3). An algorithm combines the individual ratings in each of the 6 domains to derive a global CDR score, where a CDR of 0 indicates cognitive normality and a CDR of 0.5, 1, 2, and 3 indicate questionable/very mild, mild, moderate, and severe dementia.¹⁴ The CDR has demonstrated reliability in multicenter studies.^{18 - 19} Qualitative descriptions for different levels of impairment serve as anchor points to guide ratings of function in the CA and HH domains, or “boxes,” and are shown in Table 1. For both the CA and HH boxes, “slight” impairment is rated as 0.5. Mild impairment at the 1 level for CA translates to pretense of independent function (may engage in activities and appear “normal,” although does not function without assistance) and to independence in HH for simpler activities but not more complex ones. Impairment at the CDR = 0.5 level for either CA and HH conservatively is consistent with revised MCI criteria for functional “independence,” and impairment at the CDR = 1 level for CA and HH also could be interpreted more liberally as compatible with the “mild problems” in shopping, paying bills, and meal preparation and the need for assistance allowed by the revised criteria.

Data were analyzed from the first UDS assessment for NACC participants with nonmissing data on age, sex, race, education, and CA and HH box scores and who had nonmissing data on 1 or more of the FAQ items. Participants in the cognitively normal group were required to have a CDR of 0. Eligibility criteria for the MCI group included all subtypes (amnestic MCI, memory impairment only; amnestic MCI, memory impairment plus impairment in 1 or more other domains; nonamnestic MCI, single domain; and nonamnestic MCI, multiple domains),^{8 - 9} a global CDR of 0.5, and the clinician's indication that the participant did not have dementia at that assessment. Data from participants younger than 50 years at the index assessment and individuals with Down syndrome were excluded from analyses. For persons who met eligibility criteria at more than 1 assessment (eg, a cognitively normal participant who later developed AD dementia), data from the first assessment meeting eligibility criteria were used in the analyses.

Participants were assigned to 1 of 6 groups based on their global CDR and diagnosis: CDR = 0, CDR = 0.5/MCI, CDR = 0.5/AD, CDR = 1/AD, CDR = 2/AD, and CDR = 3/AD. Two levels of FAQ ratings and CDR box scores for CA and HH that meet the revised criteria's operationalized functional “independence” for a diagnosis of MCI were examined: (1) conservative, with FAQ ratings of 0 or 1 and CDR box scores of 0 or 0.5, and (2) liberal, with FAQ ratings of 0, 1, or 2 and CDR box scores of 0, 0.5, and 1. The number and percentage of participants at each CA and HH box score level and each FAQ item rating was reported for each group.

Pairwise demographic differences between the CDR = 0, CDR = 0.5/MCI, and CDR = 0.5/AD groups were tested (t tests for continuous variables, χ² for categorical variables). Logistic regression using bias correction was used to test differences across the CDR = 0, CDR = 0.5/MCI, and CDR = 0.5/AD groups in the likelihood of functional impairment while adjusting for the effects of age, sex, education, race, and apolipoprotein E genotype. The CDR = 0.5/MCI group was used as the reference category in these analyses.

There were 17 535 individuals from 33 ADCs entered into the NACC database who met eligibility criteria, including a diagnosis of normal cognition (n = 6379), MCI (n = 4947), or probable AD (n = 6209). The mean (SD) age of the total sample was 74.6 (9.5) years (range, 50-110 years) with a mean (SD) educational attainment of 14.7 (3.5) years. Fifty-nine percent were women, 79.8% were white, and 14.9% were African American. Of the 9557 individuals for whom their apolipoprotein E allele status was known, 3994 (41.7%) were carriers of at least 1 ϵ4 allele.

Table 2 shows the demographic characteristics of the sample as a function of CDR stage and diagnosis at time of entry into the NACC database, with statistical tests of differences between the groups of greatest interest (CDR = 0 vs CDR = 0.5/MCI and CDR = 0.5/AD vs CDR = 0.5/MCI). The CDR = 0.5/MCI and the CDR = 0.5/AD groups statistically were similar with regard to age and sex. For each of the remaining demographic variables, there were significant differences between the CDR = 0, CDR = 0.5/MCI, and CDR = 0.5/AD groups. Regardless of how functional impairment was operationalized, the CDR = 0 group was less likely, and the CDR = 0.5/AD group was more likely, than the CDR = 0.5/MCI group to be functionally impaired. As expected, given the large sample sizes, all differences were statistically significant (eTable).

Table 3 shows the severity of functional impairment of the NACC sample. More than two-thirds (68.3%) of CDR = 0.5/MCI individuals were performing normally incommunity activities (CA box score = 0) vs only 23.9% of CDR = 0.5/AD individuals; for CDR = 1/AD individuals, only 0.2% were performing normally. Very similar percentages were found for normal performance of home activities (HH box score = 0) for these 3 diagnostic groups. By virtue of a score of 3 (“does with assistance”) on any of the 4 FAQ items (assembling tax records; writing checks; preparing a balanced meal; and shopping) that are equivalent to the “mild problems in paying bills, preparing a meal, or shopping” of the revised criteria for MCI,² 41.8% of CDR = 0.5/AD individuals were impaired vs 79.4% of CDR = 1/AD individuals (P < .001). When either a score of 2 or 3 for these FAQ items was considered, 73.8% of CDR = 0.5/AD individuals vs 94.7% of CDR = 1/AD individuals were impaired (P < .001).

Box scores of 0, 0.5, or 1 in CA and in HH (ie, normal, slight, and mild impairment) are allowable under the liberal operationalization of “functional independence” described by the revised criteria for MCI. The data in Table 3 indicate that 99.8% of all individuals currently diagnosed as CDR = 0.5/AD and 92.7% of all individuals currently diagnosed as CDR = 1/AD could be reclassified as having MCI with the revised criteria, because these individuals had box scores of 0, 0.5, or 1 for CA and HH (Figure 1). Limiting the operationalization of “functional independence” to box scores of 0 and 0.5 still resulted in 90.1% (for CA) and 84.5% (for HH) of currently diagnosed CDR = 0.5/AD individuals potentially meeting revised criteria for MCI, although the percentage of individuals currently diagnosed as CDR = 1/AD who could be reclassified as having MCI was notably reduced (Figure 2).

Data from FAQ items (Table 3) were consistent with the findings from the CA and HH box scores. Because “mild problems” in performing finances, cooking, and shopping or the need for assistance to perform these activities are allowed by the revised criteria for MCI, scores for the 4 corresponding FAQ items (assembling tax records, writing checks, preparing meals, and shopping) of 0 (normal), 1 (does by self with difficulty), and 2 (does with assistance) were found for between 71.5% (assembling tax records) and 94.6% (shopping) of individuals currently diagnosed as CDR = 0.5/AD and for between 31.2% (assembling tax records) to 71.7% (shopping) of individuals currently diagnosed as CDR = 1/AD (Figure 3). These individuals thus could meet the revised criteria for MCI. Limiting FAQ scores to 0 and 1 showed that between 43.4% (assembling tax records) and 80.4% (shopping) of individuals currently diagnosed as CDR = 0.5/AD and between 11.1% (assembling tax records) and 38.7% (preparing meals) of individuals currently diagnosed as CDR = 1/AD could be reclassified as having MCI (Figure 4).

The differentiation between MCI and AD dementia in its earliest symptomatic stages has been based primarily on whether the cognitive impairment interferes with the conduct of activities of daily living. Clinicians for ADC participants with ratings currently submitted to the NACC appear to follow this distinction, as two-thirds or more of individuals diagnosed with MCI are rated with no impairment in performance of activities in the community and at home, whereas more than 75% of CDR = 0.5/AD individuals have at least slight impairment in these activities. However, revised criteria for MCI² now obscure this distinction by permitting mild difficulties in functional activities to be part of the MCI spectrum. As a result, the large majority of individuals currently diagnosed with milder stages of AD dementia now could be reclassified as having MCI with the revised criteria. The elimination of the functional boundary between MCI and AD dementia means that their distinction will be based solely on the individual judgment of clinicians, resulting in nonstandard and ultimately arbitrary diagnostic approaches to MCI. This recalibration of MCI moves its focus away from the earliest stages of cognitive decline, confounds clinical trials of individuals with MCI where progression to AD dementia is an outcome, and complicates diagnostic decisions and research comparisons with legacy data.

The diagnostic overlap of MCI and AD dementia suggests ambiguity in the MCI concept,^{20 - 21} as also may be reflected by the continuing evolution of its criteria. The original criteria specified impairment in memory only with the preservation of other cognitive abilities and intact functional performance,^{6 - 7} distinguishing it from AD dementia where deficits in multiple cognitive domains and interference with functional activities are required. A multicenter clinical trial of donepezil and vitamin E in 769 individuals classified as having MCI using the original criteria found, however, that very mild impairments often were present in cognitive domains other than memory.¹¹ Criteria for MCI subsequently were modified to allow both impairment in cognitive domains other than memory and impairments in multiple domains, thus erasing one of the original distinguishing features of MCI and leaving only the “essentially normal” performance of functional activities in MCI as the difference between MCI and the milder stages of AD dementia.^{8 - 9}

Standardizing the degree of functional impairment that is associated with dementia rather than MCI has been problematic,⁸ in part because the determination of impairment is dependent on how carefully performance of usual activities is interrogated by the clinician. It nonetheless has become apparent that many individuals characterized as having MCI are functionally impaired.^{22 - 23} Ninety-six individuals with MCI in 1 study showed a wide range of impairment in everyday function as compared with 105 cognitively normal older adults,²⁴ and another study of 285 individuals with MCI found that 34.3% were restricted in their ability to perform at least 2 of 4 activities (using a telephone; handle transportation; responsibility for medications; and handle finances), compared with only 5.4% of cognitively older healthy persons.²⁵ Even using criteria for MCI that require “essentially normal” activities for daily living, as is specified for the Alzheimer's Disease Neuroimaging Initiative that examines the progression of MCI to very mild AD dementia, 104 of 283 individuals with MCI (37%) were rated as functionally impaired with 0.5 ratings for both CA and HH on the CDR or a rating of 1 for either box.²⁶

It is not surprising that their criteria essentially are indistinguishable if MCI and AD dementia are produced by the same underlying disease process. In many (but not all) instances, the underlying process for MCI is AD, a brain disorder that is characterized by continuous deterioration of synaptic and neuronal integrity.^{27 - 29} The concept of an AD continuum has been noted from many perspectives, including epidemiological,³⁰ neuropsychological,^{31 - 32} structural and metabolic imaging,³³ and pathophysiological.³⁴ The initial stages of the continuum, representing a preclinical stage of AD,⁴ are marked by no or only minimal synaptic and neuronal deterioration^{35 - 37} and correspondingly symptoms are absent (at least as detected by currently available methods). As the disease process progresses over time and neurodegeneration increases, the symptomatic stage of AD gradually develops.³⁸ The symptomatic stage may be subdivided by the degree of cognitive and functional impairment, ranging from minimal to severe (Table 4).

For virtually all persons with AD, MCI represents the earliest clinically detectable stage of symptomatic AD. Imparting binary distinctions on the cognitive continuum of AD by designating one side as “MCI” and the other as “dementia” inherently is difficult and has been made more challenging because experts cannot agree on the operationalization of MCI.³⁹ It has proven equally difficult to provide operational criteria that distinguish MCI from normal cognitive aging. Although the increased variability of cognitive performance with age⁴⁰ complicates this distinction, another factor is the reliance on cognitive performance measured at a single point to determine “impairment.” This approach, based on present level of performance, cannot capture the salient cognitive feature of AD, which is decline from that individual's previously attained abilities. Intraindividual decline (ie, the temporal loss of cognitive abilities) must be obtained with serial cognitive measurements or by a history of change from previously attained levels as reported by an observant informant.⁴¹ (Self-reported memory concerns rarely predict either impairment^{42 - 43} or future dementia,^{44 - 46} although some investigators find that subjective memory concerns in highly educated older adults may do so.⁴⁷ )

The operational cut point for MCI frequently has been a neuropsychological test score in a relevant domain, most often memory, that is 1.5 or more SDs below age-corrected means,^{48 - 51} but occasionally a score that is only 1 SD or more below normative values is used,⁵² as are other algorithms.^{53 - 54} Differing cut points produce different classifications of MCI and contribute to variable outcomes.⁵⁵ Furthermore, the effects of preclinical AD artificially lower cognitive performances in putatively “normal” samples and produce cutoffs that are too lenient to identify initial change from truly healthy cognitive aging, thus resulting in an ascertainment bias where “impairment” is detected only after notable cognitive decline has occurred.^{56 - 57} Variable cutoff scores used to characterize MCI, ascertainment bias caused by preclinical AD, and problems of misclassification when using a measurement at a single point to identify a process of intraindividual decline all introduce heterogeneity to MCI.^{49 ,58 - 62} Just as there are non-AD causes of dementia, there are non-AD causes of MCI.⁶³ The use of informant report to capture the intraindividual cognitive and functional decline that is the hallmark of symptomatic AD both minimizes this heterogeneity and enables the accurate identification of the subset of MCI that is caused by AD.^{23 ,64 - 67} The clinical diagnosis of symptomatic AD with this method, even in individuals who meet MCI criteria, is supported by autopsy confirmation of AD neuropathology at rates (92%) comparable with those found for AD dementia.⁶⁸

The revised criteria for MCI laudably recommend an etiologic diagnosis, “MCI due to AD,” when the clinical judgment is that AD is responsible for the cognitive dysfunction. The National Institute on Aging and the Alzheimer's Association work group also suggests that diagnostic confidence for AD could be enhanced by using biomarkers for amyloid-β deposition and neuronal injury.² Some clinicians, however, are reluctant to diagnose symptomatic AD in this situation, presumably because of uncertainty as to whether “MCI due to AD” will progress to AD dementia and because the diagnosis of AD is perceived as stigmatizing. As noted earlier, an informant-based diagnostic method to capture intraindividual decline accurately identifies individuals in the MCI stage who have underlying AD and who longitudinally progress in dementia severity.⁶⁷ Clinicians using other diagnostic approaches may increasingly use AD biomarkers to improve diagnostic certainty. An early etiologic diagnosis is advocated for other devastating neurodegenerative diseases such as amyotrophic lateral sclerosis^{69 - 70} without resorting to terms such as minimal motor impairment and thus allows patients and families to initiate planning to cope with the illness and enable therapeutic interventions early in the disease course. An analogous approach should be implemented for AD.

Some studies report that between 30% and 60% of physicians do not disclose the diagnosis of dementia, much less “MCI due to AD,” to their patients,^{71 - 73} although more than 94% of the same physicians disclosed a diagnosis of terminal cancer.⁷¹ Factors cited against disclosing a dementia diagnosis included concern about provoking psychological distress in patients and families.^{74 - 75} Although individuals given a diagnosis of dementia can experience catastrophic reactions, including suicide,⁷⁶ the actual risk of suicide is very small and in 1 study of community-living individuals with dementia was estimated to be less than 0.1%.⁷⁷ Patients and families have consistently indicated that receiving a diagnosis of dementia is very important to them to indicate that the cognitive symptoms have an identifiable cause and to plan for the future.^{78 - 79} A study of individuals who qualify for an MCI classification showed that depressive symptoms and measures of stress after disclosure of the diagnosis of AD dementia did not worsen from prediagnosis levels and indeed often improved for both the affected individuals and their family members.⁸⁰

The pathobiologically based “next step” now is to transition from the syndromic label of “MCI” to its etiologic diagnosis when the responsible disorder is believed to be AD.⁸¹ This “next step” circumvents the difficulty in attempting to dichotomize MCI and AD dementia and more appropriately reflects the continuum of AD. However, “MCI” still could be used to characterize individuals who may be in the early symptomatic stages of a non-AD dementing disorder, where there is less experience in accurately determining etiology than exists for “MCI due to AD,” or when potentially reversible disorders such as cognitive dysfunction associated with medications are suspected. Similar recommendations have been proposed by the International Working Group for New Research Criteria for the Diagnosis of AD. This working group proposes that “MCI” be reclassified as “prodromal AD” when this diagnosis is supported by a clinical phenotype and biomarker profile⁸² and suggests that the use of “MCI” be limited to individuals without a diagnosable illness.

The revised criteria² accept that MCI can be associated with impaired functional activities, such that the distinction of MCI from dementia now simply is a matter of an individual clinician's threshold for what represents one condition vs the other. The diagnostic overlap for MCI with milder cases of AD dementia is considerable and suggests that any distinction is artificial and arbitrary. Already, many individuals with MCI are treated with pharmacological agents approved for symptomatic AD, indicating that clinicians often do not distinguish the 2 conditions when faced with issues of medical management.^{83 - 84} It now is time to advance AD patient care and research by accepting that “MCI due to AD” is more appropriately recognized as the earliest symptomatic stage of AD.