Caselli RJ, Reiman EM, Locke DEC, Hutton ML, Hentz JG, Hoffman-Snyder C, Woodruff BK, Alexander GE, Osborne D. Cognitive Domain Decline in Healthy Apolipoprotein E ε4 Homozygotes Before the Diagnosis of Mild Cognitive Impairment. Arch Neurol. 2007;64(9):1306-1311. doi:10.1001/archneur.64.9.1306
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
Memory declines more rapidly with age in apolipoprotein E (APOE) ε4 carriers than in APOE ε4 noncarriers, and APOE ε4 homozygotes' cognitive performances correlate with stressors. These changes could represent presymptomatic disease in some, despite their youth.
To show that presymptomatic APOE ε4 homozygotes experience greater psychometric decline at a younger age than APOE ε4 heterozygotes and noncarriers before the diagnosis of mild cognitive impairment (MCI) and Alzheimer disease (AD).
Prospective observational study
Academic medical center.
A total of 43 APOE ε4 homozygotes, 59 APOE ε4 heterozygotes, and 112 APOE ε4 noncarriers aged 50 to 69 years were cognitively healthy and matched at entry according to age, educational level, and sex.
Neuropsychological battery given every 2 years.
Main Outcome Measures
Predefined test and cognitive domain decline criteria applied to consecutive epochs.
Of 214 participants, 48 showed no decline on any test, 126 showed decline on only 1 test in 1 or more domains, and 40 showed decline on 2 or more tests in 1 or more domains. Cognitive domain decline occurred in 4 of 10 APOE ε4 homozygotes 60 years and older at entry (40.0%) compared with 5 of 66 APOE ε4 heterozygotes and noncarriers (7.6%) (P = .02) and was more predictive of subsequent decline than nondomain decline (17 of 24 [70.8%] vs 29 of 70 [41.4%]; P = .01). Decline on any memory test was predictive of further decline (P < .001), as was memory domain decline (P = .006) in all genetic subgroups. Seven participants developed MCI (in 6) or AD (in 1), of whom 5 were APOE ε4 homozygotes (P = .008). Retrospective comparison showed that those who experienced multidomain, memory, and language domain decline had lower spatial and memory scores at entry than those who experienced no decline.
APOE ε4 homozygotes in their 60s have higher rates of cognitive domain decline than APOE ε4 heterozygotes or noncarriers before the diagnosis of MCI and AD, thus confirming and characterizing the existence of a pre-MCI state in this genetic subset.
Apolipoprotein E (APOE) ε4 reduces the median age at Alzheimer disease (AD) onset by gene dose from 84 years in APOE ε4 noncarriers to 68 years in APOE ε4 homozygotes.1- 6 Breitner et al,4 for example, found a disproportionately high number of APOE ε4 homozygotes in a population with AD onset between the ages of 55 and 75 years. Since 1994 we have been enrolling and following up a cohort of healthy APOE ε4 homozygotes aged 50 to 69 years at entry, demographically matched to APOE ε3/4 heterozygotes and APOE ε4 noncarriers. As a group, these APOE ε4 carriers have reduced cerebral metabolism in the same brain regions as patients with probable AD, which correlates with the APOE ε4 gene dose7 and is progressive.8 Memory declines more rapidly with age in APOE ε4 carriers than in APOE ε4 noncarriers,9 and the APOE ε4 homozygotes' cognitive performances correlate with stressors, including fatigue10 and anxiety.11 These changes could represent presymptomatic disease in some, despite their youth. We postulated, therefore, that cognitively healthy APOE ε4 homozygotes must be at greater risk for neuropsychological decline than APOE ε4 heterozygotes and noncarriers preceding clinical conversion to mild cognitive impairment (MCI) and AD as they age into their 60s.
We tested the hypothesis that there is a higher rate of neuropsychological decline among APOE ε4 homozygotes than APOE ε4 heterozygotes and noncarriers during the seventh decade of life. Clinical criteria for MCI12,13 and AD14 emphasize but do not operationally define cognitive domain decline, so we created objective criteria for test score, test (some tests have multiple scores), and cognitive domain decline (a decline on at least 2 tests of that cognitive domain). We analyzed 2 age-decile groups (sixth and seventh decades) for emergent patterns of decline and the influence of APOE on such patterns.
Participants were 50 to 69 years old at entry and were identified through local newspaper advertisements that requested healthy individuals who had a first-degree relative with AD.9 Genetic determination of APOE allelic status was performed using a polymerase chain reaction–based assay.15
Each APOE ε4 homozygote was matched to 1 APOE ε4 heterozygote (APOE ε3/4 genotype) and 2 APOE ε4 noncarriers for sex, age (within 3 years), and educational level (within 2 years). Screening tests included medical history, neurologic examination, Folstein Mini-Mental State Examination, Hamilton Depression Rating Scale, and the Structured Psychiatric Interview for Diagnostic and Statistical Manual of Mental Disorders (Third Edition Revised).16 None met criteria for MCI,12,13 AD,14 any other form of dementia, or major depressive disorder.16 Those who completed at least 2 epochs of testing were included.
A 4-hour battery of neuropsychological tests17 was administered approximately every 2 years. Test-retest changes were calculated for each score on each test between each pair of consecutive epochs. We assumed that disease-related decline would manifest as interepoch decline, although we realize that longer decline intervals are possible. Decline between consecutive epochs that exceeded 2 SDs beyond decline of the entire cohort was defined as abnormal.
For tests with multiple scores, only a single test score was required to define test decline. For example, only 1 of the 3 scores used from the Auditory Verbal Learning Test (AVLT)17 (short-term delayed recall, long-term delayed recall, and percent recall) was required for a diagnosis of AVLT decline.
Five cognitive domains were defined. Four reflected intellectual skills: executive, memory, language, and spatial. The fifth was a behavioral domain, since depression and paranoia frequently occur in early AD. Tests that comprised these domains were as follows:
Executive: Wechsler Adult Intelligence Scale–Revised (WAIS-R) Freedom From Distractibility, Controlled Oral Word Association Test, Wisconsin Card Sorting Test (categories, total errors, and perseverative errors), and the Paced Auditory Serial Attention Task (3- and 2-second administration).
Memory: AVLT (short-term delayed recall, long-term delayed recall, and percent recall), Selective Reminding Test (free and cued recall), Complex Figure Test recall, and Visual Retention Test (total correct).
Language: WAIS-R Vocabulary (age scaled), WAIS-R Similarities (age scaled), Boston Naming Test, and Token Test.
Spatial: WAIS-R Block Design (age scaled), Complex Figure Test copy score, Facial Recognition Test, Judgment of Line Orientation Test.
Behavioral: Hamilton Depression Scale, Beck Depression Inventory, Personality Assessment Inventory Clinical Scale for Depression, Personality Assessment Inventory Clinical Scale for Paranoia.
Additionally, all received the WAIS-R Information and Digit Symbol Substitution, Mattis Dementia Rating Scale, Wechsler Memory Scale Orientation, and the Functional Assessment and Instrumental Activities of Daily Living Questionnaires. Subjective assessment of cognitive and behavioral change during the preceding year was obtained from a close observer (typically a spouse) and the participant using previously unpublished questionnaires.
Cognitive domain decline was defined as a decline on 2 or more tests within that domain. From these definitions, the following 5 decline categories were created:
No decline on any test score
Decline on 1 test score (in 1 domain)
Decline on 1 test score in 2 or more domains
Single-domain decline (without or with single-test decline in 1 or more other domains)
Multidomain decline (without or with single-test decline in 1 or more other domains)
Comparisons between groups for normally distributed data were made using either analysis of variance or 2-sample t tests for continuous data and the χ2 or Fisher exact test for categorical data. The Kruskal-Wallis or Wilcoxon rank sum test was used for scores that were not normally distributed. All reported P values are 2-tailed. Participants who were evaluated for cognitive difficulties arising any time after at least 1 epoch of testing and who fulfilled published criteria for MCI12,13 or AD14 were considered clinical converters.
Of 268 participants who enrolled, 214 completed at least 2 epochs of testing with a mean ± SD duration of observation of 56.7 ± 30.1 months (median, 53.0 months) and were included in this analysis. Compared with participants, dropouts were less likely to report a family history of dementia in a first-degree relative (P < .001, χ2 test), but no differences were found in age, sex, educational level, or score on any psychometric measure.
From the entry epoch, the mean ± SD duration was 23.9 ± 5.2 months to epoch 2, 51.2 ± 11.2 months to epoch 3, 74.9 ± 8.4 months to epoch 4, 96.8 ± 9.9 months to epoch 5, 110.5 ± 12.9 months to epoch 6, and 125.8 ± 1.3 months to epoch 7. Table 1 summarizes the demographic data of the participants. There were 43 APOE ε4 homozygotes, including 33 aged 50 through 59 years and 10 aged 60 through 69 years at entry. The mean ± SD Mini-Mental State Examination score at entry for the entire cohort was 29.6 ± 0.7, which did not differ among the 3 genetic subgroups for those aged 50 through 59 years (P = .37) or 60 through 69 years (P = .29; analysis of variance). The frequency of a first-degree relative was slightly lower among the APOE ε4 noncarriers (who otherwise had a second-degree relative with dementia).
Table 2 summarizes the categories of decline. Forty-eight of 214 participants showed no decline on any test (category 1), and this finding did not differ by APOE ε4 genotype (95% confidence interval [CI], −0.01 to 0.26). A total of 126 showed decline on only 1 test in 1 (n = 63, category 2) or more domains (n = 63, category 3), and 40 showed decline on 2 or more tests in 1 (n = 35, category 4) or multiple domains (n = 5, category 5).
Single-domain decline (category 4) affected memory in 14, behavior in 8, executive skills in 7, and language in 6 participants. Those who experienced executive and language decline were younger than the other single-domain groups (P < .001; 2-sample t tests). None had single-domain spatial decline. Five had multidomain decline (category 5) that involved spatial in 4, executive skills and memory each in 3, and language in 2.
Among those aged 60 through 69 years, homozygotes had a higher proportion of cognitive domain decline than heterozygotes and noncarriers (4 of 10 [40.0%] vs 5 of 66 [7.6%]; P = .02; Fisher exact test). Cognitive domain decline occurred more frequently in the older homozygotes group than in the other older genetic subgroups in the spatial domain (2 of 10 [20.0%] vs 0 of 10 [0%]; P = .02; Fisher exact test) within a multidomain context, but in the younger homozygotes subgroup, domain decline occurred more frequently in the behavioral domain (5 of 33 [15.2%] vs 3 of 105 [2.8%]; P = .02; Fisher exact test). Memory domain decline occurred in 6 of 43 (14.0%) of the homozygotes group overall (single domain only in 4 of 43 [9.3%]) and occurred more frequently in both younger (5 of 33 homozygotes [15.2%] vs 9 of 105 other older genetic [8.6%]) and older (1 of 10 homozygotes [10.0%] vs 2 of 66 other older genetic [3.0%]) subgroups, but too few homozygotes were available to assess differences among genetic subgroups (95% CI, −0.02 to 0.21). Multidomain decline occurred more frequently in older homozygotes (2 of 10 [20.0%] vs 0 of 66 [0%]; P = .02; Fisher exact test) and younger homozygotes (4 of 33 [12.1%] vs 3 of 105 [2.8%]; P = .06; Fisher exact test) compared with the other genetic subgroups.
Subsequent epochs of observation were available in 94 of the 166 individuals who experienced decline on at least 1 test (Table 3). A total of 36 of 94 (38.3%) showed a mixed pattern of both decline and improvement, and 35 of 94 (37.2%) showed neither. Only 8 of 94 (8.5%) declined alone, and only 15 of 94 (16.0%) improved alone. Patterns differed according to age group (P = .01) but not APOE status (P = .17; χ2 test). Further decline occurred in 17 of 24 of those with domain-specific decline (70.8%) and 29 of 70 of those with decline that was not domain specific (41.4%) (P = .01; χ2 test). Any level of decline on any memory test correlated with subsequent decline (P < .001; χ2 test), as did memory domain decline (P < .001; χ2 test) for the cohort overall as well as for each of the individual genetic subgroups: noncarriers, 13 of 22 (59.1%) (P = .01); APOE ε4 heterozygotes, 11 of 15 (73.3%) (P = .02); and APOE ε4 homozygotes, 8 of 10 (80.0%) (P = .02). Other test or domain declines were not correlated with subsequent decline.
Six participants developed MCI (amnestic in 5 and visual in 1), and 1 developed AD. Among the 6 patients with MCI, 3 progressed to AD (autopsy obtained with AD confirmation in 1 to date). All were in their 60s at the time of symptomatic conversion. Mean ± SD time to clinical presentation after entry was 54.4 ± 35.7 months, and mean ± SD age at diagnosis was 64.3 ± 3.7 years. Five were APOE ε4 homozygotes (11.9% of all the APOE ε4 homozygotes enrolled), 1 was an APOE ε4 heterozygote (2.0% of APOE ε4 heterozygotes enrolled), and 1 was an APOE ε4 noncarrier (1.0% of APOE ε4 noncarriers enrolled) (P = .008; Fisher exact test). Stated differently, 5 of 7 converters (71.4%) were homozygotes even though homozygotes constituted only 20.0% of the cohort (43/214).
Baseline performances were compared between those who did not decline and the other 4 categories of decline (Table 4). Compared with those who did not decline (category 1), few differences occurred overall, reconfirming the truly incident nature of observed decline. Those who experienced multidomain, language, and memory decline had lower scores on spatially oriented tests that encompassed memory and spatial domains. Those who experienced multidomain and language decline additionally had lower scores on the Boston Naming Test. Mean scores for all tests at entry were nonetheless well within normal limits.
Our results show that by the time they reach their early 60s, many APOE ε4 homozygotes experience cognitive domain decline that precedes the clinical diagnosis of MCI or AD. The number of clinical converters in this study was small, but they also were mostly APOE ε4 homozygotes. Our findings are consistent with the original formulation of Corder et al1 and with previous case-control studies,1,2 epidemiologic studies,3,6 and meta-analyses5 that have suggested a gene dose–related earlier age at AD onset.
Memory was normal at entry, but decline on the incident memory test predicted subsequent further decline. Baseline performances in those who declined on the multidomain, memory, and language tests were characterized by lower memory and spatial domain performance than those who did not decline, a neuropsychological pattern that typifies AD and suggests that AD may have already started in these individuals despite their normal entry test scores. Because decline was detected in individuals and not simply at a group level, our data point out the potential utility of longitudinal change as a clinically meaningful harbinger of impending symptomatic conversion when baseline neuropsychological scores fail to reveal impairment.
Before addressing our results further, the limitations of this study must be weighed. Although this study capitalized on rigorously defined, prospectively tested decline categories, nearly every one of our decisions in creating these definitions can be questioned: the 2-SD threshold for abnormal decline, use of a 2-year interepoch change, choice of tests, choice of test scores, choice of domains, and the categories of decline. The number of clinical converters to date is small. Even so, our definitions of longitudinal cognitive decline were more sensitive than cross-sectional measures in predicting further decline or clinical conversion, and they supported our prediction that homozygotes who are at the highest risk for AD would experience the greatest decline preclinically as well as clinical conversion to MCI and AD. Our ongoing longitudinal study will determine the extent to which our baseline measurements and longitudinal declines predict conversion in larger samples. Studies in subsets of these individuals may permit us to determine the extent to which these declines predict rates of conversion in bioimaging and other biomarker measurements of AD progression and ultimately the extent to which these declines predict rates of conversion to the neuropathologic diagnosis of AD.
Many longitudinal aging studies are currently being performed in the United States and elsewhere, and an approach to the rational interpretation of longitudinal decline that precedes symptomatic diagnosis is needed, which is underscored by our observed high frequency of test score decline. Ours is one such approach. Presymptomatic individuals with neuropsychologically documented longitudinal decline clearly exist but likely constitute different groups, as shown by our results. Further supporting this concept, Storandt et al18 recently compared rates of progression of 3 different groups of individuals at an early symptomatic stage of disease using a Clinical Dementia Rating of 0.5 to 1.0: those who met original amnestic MCI criteria,12 those who met revised criteria recognizing nonmemory domain decline,13 and those who did not meet either set of criteria. Correlation with neuropathologically confirmed AD was high in all 3 groups, but the rate of progression was slower in the last group, showing that it may have been a more mildly affected group.18 By comparison, our APOE ε4 homozygotes with domain-specific neuropsychological decline were asymptomatic at entry, showed neuropsychological decline that preceded clinical presentation, and may represent an even earlier, pre-MCI stage of disease. Given the recent redefinition of MCI to encompass different cognitive domains, we regard the preclinical categorization of cognitive domain–specific decline to be a particular strength of our approach and the subgroups identified a rich resource for further study.
In conclusion, presymptomatic APOE ε4 homozygotes experience a high rate of cognitive domain decline by the time they reach their 60s. Incident memory decline in these and other apparently healthy individuals is also strongly predictive of subsequent decline. Ongoing study of this important cohort will provide further insight into the clinical and neurobiological significance of pre-MCI syndromes consisting of neuropsychologically defined decline in specific cognitive domains and its relative predictive value for incipient conversion to MCI, AD, and related forms of dementia.
Correspondence: Richard J. Caselli, MD, Department of Neurology, Mayo Clinic, 13400 E Shea Blvd, Scottsdale, AZ 85259 (email@example.com).
Accepted for Publication: January 22, 2007.
Author Contributions: Drs Caselli, Reiman, Locke, and Osborne and Mr Hentz had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Caselli, Reiman, and Alexander. Acquisition of data: Caselli, Reiman, Hutton, Hoffman-Snyder, Woodruff, and Osborne. Analysis and interpretation of data: Caselli, Reiman, Locke, Hentz, Woodruff, Alexander, and Osborne. Drafting of the manuscript: Caselli and Hutton. Critical revision of the manuscript for important intellectual content: Caselli, Reiman, Locke, Hentz, Hoffman-Snyder, Woodruff, Alexander, and Osborne. Statistical analysis: Hentz and Alexander. Obtained funding: Caselli and Reiman. Administrative, technical, and material support: Reiman, Locke, Hutton, Woodruff, and Osborne.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grants RO1 MH57899-01A1 (Drs Reiman and Caselli) and P30 AG19610-01 (Dr Reiman) and the Arizona Alzheimer's Disease Research Consortium (Dr Caselli).
Additional Contributions: Sandra Yee-Benedetto, BA, Bruce Henslin, BA, Jessie Jacobsen, BA, Jennifer Adamson, MBA, BS, Ashley Cannon, BA, Jennifer Gass, BS, Jennifer Tessier, BA, Brie Noble, BA, Anita Prouty, BS, and Christine Burns, BA, provided excellent technical support.