Biracial Population Study of Mortality in Mild Cognitive Impairment and AD

Participants

Subjects are from the Chicago Health and Aging Project, a longitudinal study of aging and AD conducted in 3 adjacent neighborhoods in Chicago [18]. It began with a census of the community, after which all those aged 65 years or older were invited to participate in an in home interview that included brief tests of cognitive function. A stratified random sample of those interviewed was selected for a detailed clinical evaluation (see below). Approximately 3 years later, the population interview was repeated and a stratified random sample of those determined to be without dementia in the previous wave of data collection had a clinical evaluation. These two steps (interview of population, clinical evaluation of a new sample of persons previously deemed to be free of dementia) have been repeated at intervals of approximately 3 years, with the fifth wave of data collection now in progress. For these analyses, we included all persons who completed a clinical evaluation in the second wave of data collection or in a subsequent wave (i.e., third, fourth, or fifth) and monitored vital status after the date of the clinical evaluation. In the event that a person was sampled for more than one clinical evaluation, we used the earliest to maximize the length of the monitoring period. Those selected had a mean age of 80.1 (SD = 5.9) and had completed a mean of 12.7 years of education (SD = 3.6); 61.3% were women and 52.5% were African American.

Clinical Evaluation

Each participant had a structured uniform clinical evaluation that included a medical history, complete neurological examination, and cognitive function testing (see below), as previously described [19,20]. On the basis of this evaluation and an in person examination of the subject, an experienced physician classified individuals with respect to dementia and AD using the guidelines of the joint working group of the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer's Disease and Related Disorders Association [21]. The criteria require a history of cognitive decline and impairment in at least two cognitive domains. For the diagnosis of AD, one of the impaired domains must be memory.

MCI was diagnosed in a 2-step process, as previously described [22,23]. First, after review of all cognitive data, a neuropsychologist rated impairment in 5 cognitive domains. To maintain consistency in ratings, educationally adjusted cutoff scores on 11 tests and algorithmic ratings of each domain were provided [22]. Second, the clinician determined whether dementia criteria were met. Those with cognitive impairment but without dementia were deemed to have MCI. In other cohorts, persons meeting these criteria have shown intermediate levels of cognitive decline, AD pathology, and cerebral infarction relative to dementia and no cognitive impairment subgroups [24].

Apolipoprotein E genotyping was done by persons blinded to all clinical data using methods adapted from Hixon and Vernier [25].

Assessment of Cognitive Function

In an approximately 1-hour session, cognitive function was assessed with a battery of 18 tests. The Mini-Mental State Examination was used to describe level of global cognition but was not used in analyses. Episodic memory was assessed with 7 measures: immediate and delayed recall of the East Boston Story and story A from the Wechsler Memory Scale-Revised and Word List Memory, Recall, and Recognition. Semantic memory was assessed with a 15-item version of the Boston Naming Test, Verbal Fluency, and a 15-item version of the National Adult Reading Test. Digit Span Forward, Digit Span Backward, and Digit Ordering were used to measure working memory. Perceptual speed was measured with the oral version of the Symbol Digit Modalities Test and Number Comparison, and visuospatial ability was assessed with a 15-item version of Judgment of Line Orientation and a 12-item version of Standard Progressive Matrices.

To minimize floor and ceiling artifacts and other sources of measurement error, we used composite measures of two or more tests in analyses. A composite measure of global cognition was constructed using all 17 tests. To develop measures of specific cognitive functions, we hypothesized that the tests could be grouped into 5 domains: episodic memory, semantic memory, working memory, perceptual speed, visuospatial ability. In participants without dementia, we performed a factor analysis of the 17 tests with varimax rotation. The factor analytic grouping showed good agreement with the hypothesized grouping (Rand's statistic [26], a measure of goodness of fit ranging from -1 to 1, was 0.62, p<.001). Therefore, we used the hypothesized groupings to make summary measures of episodic memory (7 tests), semantic memory (3 tests), working memory (3 tests), perceptual speed (2 tests), and visuospatial ability (2 tests). For all composite scores, raw scores were converted to z scores, using the mean and standard deviation of the full group, and z scores on component tests were averaged to yield the composite, as previously described for other cohorts [27,28].

Determination of Vital Status

Information about death was obtained through multiple sources as previously reported [29]. Study personnel often learned of the death of a participant when attempting to schedule follow-up interviews. In addition, we regularly scanned local newspapers for obituaries and searched web sites such as www.ancestry.com. We also attempted to verify all deaths by acquiring a death record from the National Death Index.

Data Analysis

We analyzed the relation of diagnosis to risk of death in a series of proportional hazards models [30]. In these analyses, persons without cognitive impairment were used as a reference group that was contrasted with each diagnostic group. Terms were also included to control for age, sex, education, and race. In subsequent analyses, we added terms for comorbid medical conditions, possession of a copy of the apolipoprotein E ε4 allele, or socioeconomic status. We also conducted separate analyses in white and African American persons and repeated the original analysis with terms for the interaction of race with MCI and with AD.

In separate analyses within the MCI and AD subgroups, we examined the relation of severity of cognitive impairment to risk of death. The initial analyses included a term for level of global cognition at baseline. We then conducted separate analyses in black and white persons and repeated the initial analysis of the AD group with a term for the interaction of race with global cognitive score. In subsequent analyses, we substituted measures of specific cognitive functions for the global measure, first in separate models and then together in a single model.

Models were graphically and analytically validated. Programming was done in SAS [31].

Dementia Diagnosis and Mortality

Vital status was monitored for up to 10 years (mean = 4.7, SD = 3.0). During this period, 634 individuals died (37.0%), including 25.8% of those without cognitive impairment, 40.4% of those with MCI, 59.1% of those with AD, and 60.0% of those with other forms of dementia. We examined the relation of diagnosis to risk of death in a series of proportional hazards models.

These and all subsequent analyses were adjusted for the potentially confounding influences of age, sex, race, and education. Individuals without cognitive impairment served as a reference group that was contrasted with each of the other diagnostic subgroups (i.e., MCI, AD, other dementia). As shown in the first column of Table 2, compared to those without cognitive impairment, risk of death was about 50% higher in those with MCI and nearly three times higher in those with AD or other forms of dementia. Results were similar after controlling for common chronic medical conditions (i.e., heart disease, hypertension, stroke, diabetes, cancer; hazard ratio [HR] for MCI = 1.44; 95% confidence interval [CI]:1.19,1.75; HR for AD = 2.81; 95% CI: 2.25, 3.50) possession of at least one apolipoprotein E ε4 allele (HR for MCI = 1.45; 95% CI:1.19, 1.77; HR for AD = 2.91; 95% CI: 2.32, 3.65), or socioeconomic indicators (i.e., occupation and income; HR for MCI = 1.49; 95% CI: 1.21, 1.82; HR for AD = 2.81; 95% CI:2.233, 3.55).

Because little is known about racial differences in survival in AD, we conducted separate analyses for African American and white persons. In these analyses, the associations of MCI and AD with risk of death in African Americans (second column of Table 2 and upper panel of Figure 1) appeared slightly weaker than the same associations in whites (third column of Table 2 and lower panel of Figure 1). This apparent difference was not significant, however: in an analysis that included terms for the interactions of race with the indicators for MCI and AD, neither interaction was significant (for race × MCI, estimate = -0.29, SE = 0.19, p =.13; for race × AD, estimate = -0.36. SE = 0.21, p = .08).

Cognitive Impairment and Mortality Within Diagnostic Subgroups

MCI and AD are clinically heterogeneous conditions. One important way in which affected persons differ is in overall severity of cognitive impairment. To determine whether individual differences in the severity of these conditions added to the prediction of mortality, we conducted separate analyses within the MCI and AD subgroups. Each analysis used baseline score on the composite measure of global cognition (mean = 0.35, SD = 0.59 in full cohort) as an indicator of severity of cognitive impairment. Lower level of global cognition was associated with increased risk of death in both the MCI (HR = 2.65; 95% CI: 1.68, 4.16) and AD (HR = 2.29; 95% CI: 1.65, 3.19) subgroups. Thus, in MCI a relatively low global cognitive score (score = 0.01, 25^th percentile in MCI group) was associated with a 58% increase in risk of death compared to a relatively high score (score = 0.48, 75^th percentile in MCI group). Risk of death in AD was increased by about 58% for a low score (-0.70) versus a high score (-0.15). To determine whether these effects varied by race, we conducted separate analyses in black and white subjects. In MCI, the correlation of severity of cognitive impairment with mortality was equivalent in African Americans (HR = 2.54; 95% CI: 1.31, 4.94) and whites (HR = 2.44; 95% CI: 1.30, 4.57). In AD, severity of cognitive impairment had a slightly stronger association with mortality in black (HR = 3.38; 95% CI: 2.06, 5.53) than in white participants (HR = 1.62; 95% CI: 1.01,2.59), confirmed by an interaction between race and global cognitive score in a subsequent analysis (estimate = -0.64, SE = 0.31, p = .04).

Persons with MCI and AD also vary in the cognitive domains that are most compromised. Therefore, we conducted a final series of analyses to assess whether the association of severity of cognitive impairment with mortality in these conditions varied across domains of cognition. Within each diagnostic subgroup, we related baseline level of function in each of 5 cognitive domains to mortality, first with a separate model for each domain (Table 3, left columns) and then with all 5 domain scores simultaneously analyzed (Table 3, right columns). In the separate models, lower levels of function in nearly all cognitive domains were associated with increased risk of death in both MCI and AD, consistent with a global severity effect. By contrast, in the simultaneous models, only perceptual speed was related to increased mortality in both MCI and AD, with no associations in the other cognitive domains, suggesting a much more selective effect.