Association Between Apolipoprotein Eϵ4 and the Rate of Cognitive Decline in Community-Dwelling Elderly Individuals With and Without Dementia

Objective  To determine whether the apolipoprotein E ϵ4 allele (apoE ϵ4) is associated with cognitive decline in individuals with and without dementia, we conducted a 4-year longitudinal study of subjects with a range of cognitive function.

Setting  At baseline, respondents (n=511) were randomly selected according to age and Mini-Mental State Examination score from a community-based study of dementia among noninstitutionalized persons aged 65 to 84 years. Respondents were examined at baseline and followed up in 3 annual visits. At baseline, subjects were classified as having normal cognitive function, minimal dementia, or dementia, according to criteria from the Cambridge Examination for Mental Disorders of the Elderly (CAMDEX) and the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition . Of the 511 respondents at baseline, 405 who were examined at least 2 times are included in this analysis.

Main Outcome Measures  Cognitive decline was determined by a slope estimating yearly change in score on the neuropsychological test, the CAMCOG (the cognitive section of the CAMDEX), and its subscales of memory and nonmemory functions.

Results  Among the subjects who had normal cognitive function at baseline, apoE ϵ4 carriers showed a significantly greater decline (P<.001) in score on the CAMCOG compared with noncarriers. Differences in decline on the memory and nonmemory subtests were also significant (P<.001). Rates of cognitive decline were not related to apoE ϵ4 status in the groups with minimal dementia and dementia.

Conclusions  In our community-based sample, apoE ϵ4 was associated with the rate of cognitive decline prior to the clinically symptomatic phase of dementia. Knowing the apoE ϵ4 status of those already symptomatic for dementia may not improve knowledge about a patient's prognosis.

ALTHOUGH the apolipoprotein E ϵ4 allele (apoE ϵ4) has been repeatedly demonstrated to be a risk factor for sporadic and familial Alzheimer disease (AD),^1-3 the mechanisms by which the allele may alter the risk for dementia are not well understood. One indication of how apoE ϵ4 may be involved in the disease process is the extent to which the allele is associated with or prognostic for disease progression in the subclinical and clinical stages of AD.

Three diagnostic groups have formed the basis of studies on the progression of cognitive decline and dementia in relation to apoE ϵ4: individuals with normal cognitive function, those with mild cognitive impairment, and those with clinically evident dementia. The few population-based studies of community-dwelling elderly individuals suggest that apoE ϵ4 is associated with a greater risk for cognitive decline and dementia.^4-6 Studies of subjects with minimal dementia and dementia have been mostly clinic based; some studies, in particular those of individuals with minimal dementia, show an association between apoE ϵ4 and the progression of dementia,^7-9 while others do not.^10-16 These conflicting results might be related to selection biases or to differences in the stages of the disease included in the various clinical samples. This makes comparisons across diagnostic subgroups difficult.

In this study, we compared rates of cognitive decline with apoE ϵ4 status in subjects with normal cognitive function, minimal dementia, and dementia. All subjects were selected from the same source population of community-dwelling persons aged 65 to 84 years. We also investigated whether the apoE ϵ4 allele was differentially associated with a decline in memory and nonmemory functions, as has previously been suggested.¹⁷

Subjects were participants in the Amsterdam Study of the Elderly, a 2-phase study on cognitive decline and dementia in community-dwelling elderly individuals. The design and sampling methods of the study are described in more detail elsewhere.^18,19 Briefly, a random age-stratified (65-84 years in 5-year strata) sample was selected from the lists of 30 general medical practices in the municipality of Amsterdam. A total of 4051 subjects (71.5% response rate) participated in this cross-sectional study. Subjects were visited at home by trained lay interviewers, who administered, among other items, the Mini-Mental State Examination (MMSE).²⁰

From the sample of 4051 subjects, we excluded those (n=95) who were nearly blind, deaf, or did not have a thorough command of the Dutch language. From the remaining 3956, we invited for follow-up all respondents (n=156) who scored less than 22 points on the MMSE, a random age-stratified sample of approximately 45% (n=416) of subjects with borderline (22-26 points) scores on the MMSE, and a random age-stratified sample of approximately 7% (n=215) of subjects with good (27-30 points) MMSE scores. We oversampled those with borderline cognitive function, since 1 objective of the follow-up study was to better characterize individuals with mild cognitive impairment. Of the total 787 subjects who were invited, 511 participated in the follow-up study. Compared with the respondents, the nonrespondents to the follow-up study did not differ in age, sex distribution, MMSE score, or years of education.

The respondents were examined at baseline and then visited annually for 3 years. During these visits, the diagnostic procedure for dementia was administered. Blood was drawn at baseline. There were 485 subjects with complete baseline data and 405 subjects with complete data from at least 2 examinations.

Dropout in the follow-up study did not affect the proportion of subjects with the ϵ4 allele, which was 30.6% (n=148) (SE, 2.3%) in the total sample at baseline and 28.0% (n=71) (SE, 3.0%) in the group of 254 subjects who participated through the end of the study. In addition, the proportion of subjects with the apoE ϵ4 allele did not change within the 3 diagnostic groups. Dropouts did not differ by sex or years of education, although they were significantly older (78.4 years [SE, 4.8 years] vs 75.7 [SE, 5.7 years]; P <.001). Although the individuals without dementia at baseline had proportionately more visits (mean±SD visits for subjects without dementia, with minimal dementia, and dementia were 3.5±0.8, 3.3±0.8, and 3.1±0.7, respectively; P =.005), within the diagnostic groups there were no differences in the number of visits according to apoE ϵ4 status. The analytic sample for the study of cognitive decline included the 405 subjects who had at least 2 scores on the CAMCOG.

At each visit, a physician and nurse from the study assessed dementia with the protocol of the Cambridge Examination for Mental Disorders of the Elderly (CAMDEX).²¹ The CAMDEX includes an interview with the participant and, if possible, an informant, a physical examination, and a cognitive section (the CAMCOG). The diagnoses of dementia and its subtypes were made in conference under the supervision of a senior psychiatrist and neurologist. Diagnoses for dementia were based on criteria from the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition,²² and diagnoses of AD were based on the NINCDS-ADRDA (National Institute of Neurological Disorders and Stroke–Alzheimer's Disease and Related Disorders Association).²³ For these analyses, respondents were classified as having normal cognitive function, minimal dementia (according to CAMDEX), or dementia. The CAMDEX category of minimal dementia includes subjects who show unmistakable cognitive symptoms (mostly amnestic symptoms) but do not yet satisfy the criteria for dementia in the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition. The diagnosis of minimal dementia is comparable to a Clinical Dementia Rating²⁴ of 0.5 (questionable dementia).

Cognitive function and decline were measured with the CAMCOG. This is a 60-item omnibus test of orientation in time and place, learning, recent and remote memory, attention, visual and tactile perception, praxis, calculation, receptive and expressive language functions, and verbal abstraction. The CAMCOG includes the MMSE.²⁰ The maximum CAMCOG score is 107 points. For the present analysis, the CAMCOG was divided into memory and nonmemory subscales. The memory subscale includes questions about orientation, learning, and memory; the nonmemory subscale includes the remaining questions on language function, attention, praxis, calculation, verbal abstraction, and perception.

In the first year, blood was sampled and plasma was stored at −80°C. Apolipoprotein E phenotypes were determined by isoelectric focusing of delipidated plasma samples, followed by immunoblotting.²⁵

We included in the group without apoE ϵ4 those with the phenotypes ϵ2/ϵ2, ϵ2/ϵ3, and ϵ3/ϵ3; carriers included those with at least 1 ϵ4 allele, ie, with the phenotypes ϵ2/ϵ4, ϵ3/ϵ4, or ϵ4/ϵ4. The cross-sectional comparisons of demographic variables, MMSE scores, CAMCOG scores, and apoE phenotypes by diagnostic group were done with analysis of variance for continuous variables and the χ² test for categorical variables. Age-adjusted odds ratios (95% confidence interval [CI]) for apoE ϵ4 allele dose and prevalent and incident dementia were calculated with logistic regression. The relation of cognitive decline to apoE ϵ4 status and diagnostic group was examined using multiple regression.

Cognitive decline was expressed as a slope and calculated for each individual using a linear model with the visit number (1-4) as the independent variable and the cognitive score obtained at each visit as the dependent variable. Separate slopes were calculated for the total CAMCOG score, memory subscore, and nonmemory subscore. Each subject's slope was then entered as the dependent variable into the separate linear regression models. Controlling for years of education and age at baseline, we then used analysis of variance to examine the difference in slope by apoE status, diagnostic group, and apoE ϵ4 allele status within the diagnostic groups. The analysis of the slope by apoE ϵ4 allele status within the diagnostic group was used to test whether the relation between apoE ϵ4 status and cognitive decline differed among the 3 diagnostic groups. We first tested this difference for statistical significance by entering into the model an interaction term, which is the product of a diagnostic group multiplied by apoE status. We then compared in the separate strata of each diagnostic group the difference in the slope of the cognitive test score by apoE status. The above comparisons were also examined with a random effects model that accounts for unequal numbers of observations per person over time.²⁶ Both methods resulted in the same conclusions concerning the importance of the interaction, ie, concerning the differences in cognitive decline by apoE status across diagnostic group.

The diagnostic groups differed significantly in demographic characteristics and cognitive test variables at baseline (Table 1). The subgroups with dementia and minimal dementia included more women, older subjects, and those with slightly fewer years of education than those in the subgroup with normal cognitive function. Individuals in the group with dementia had the lowest cognitive scores, and those in the group with minimal dementia had scores in between those of individuals with and without dementia. The proportion of apoE ϵ4 allele carriers was significantly higher in the subgroups with minimal dementia and dementia compared with the subgroup with normal cognitive function.

Most (74.4% [n=32]) of the 43 patients with dementia had been diagnosed as having AD, and 62.5% (n=20) of patients with AD were ϵ4 allele carriers, compared with only 9.1% (n=1) of patients with other types of dementia (P =.002).

The age-adjusted odds ratio for prevalent dementia associated with the ϵ4 allele was 2.65 (95% CI, 1.53-4.60). During the course of the study, 34 subjects were diagnosed as having dementia, 8 from the group with normal cognitive function at baseline and 26 from the group with minimal dementia. The age-adjusted odds ratio for incident dementia associated with an apoE ϵ4 allele was 2.66 (95% CI, 1.27-5.58). Among subjects with normal cognitive function at baseline, the age-adjusted odds ratio for incident dementia associated with apoE ϵ4 was 1.89 (95% CI, 0.43-4.37). Among those with minimal dementia, the age-adjusted odds ratio for incident dementia was 3.31 (95% CI, 0.94-11.59). The age-adjusted odds ratio associated with apoE ϵ4 for patients with incident and prevalent AD combined was 3.62 (95% CI, 1.54-8.53).

In the overall analysis, the rate of decline in all 3 cognitive scores differed significantly by diagnostic subgroup, irrespective of ϵ4 status (P<.001).

Table 2 shows the annual decline in the 3 cognitive scores of ϵ4 allele carriers and noncarriers within the 3 diagnostic subgroups. The interaction between apoE status and diagnostic group was significant for the total CAMCOG (P =.002). This significant interaction reflected the finding that in the group with normal cognitive function, the ϵ4 allele carriers declined significantly more than non-ϵ4 carriers. There remained a significant decline in cognitive score among the group with normal cognitive function even after all patients with incident minimal (n=35) and clinical (n=8) dementia were excluded from the analysis (the mean decline in score for ϵ4 carriers and noncarriers was −1.35 [95% CI, −1.79 to 0.91] and −0.40 [95% CI, 0.64 to −0.16], respectively).

In contrast to the findings for subjects with normal cognitive function, there were no significant differences in the declines of cognitive test scores between ϵ4 allele carriers and noncarriers in the minimal dementia and dementia subgroups. The results were similar when all patients with dementia but without AD were removed from the group with dementia. When all patients with prevalent (n=27) and incident (n=25) AD were grouped (n=52), there was a nonsignificant difference between carriers and noncarriers (P>.05). The rate of decline in the non-ϵ4 carrier group with AD was −4.09 (95% CI, −4.76 to −3.41), and that in the ϵ4 carrier group with AD was −4.36 (95% CI, −4.94 to −3.77).

The same pattern seen with the analyses of the total CAMCOG scores was evident for the memory (P =.03) and nonmemory (P =.02) subscales. Among the subjects with normal cognitive function, decline was significantly greater in both the memory and nonmemory sections of the CAMCOG (Table 2). There was no difference in decline by ϵ4 status within the minimal dementia and dementia groups.

We found that during an average period of 3 years, apoE ϵ4 allele carriers who had normal cognitive function at baseline had a significantly greater cognitive decline compared with noncarriers.

In contrast, we did not find a faster cognitive decline in symptomatic ϵ4 allele carriers compared with noncarriers. In these analyses, the association of apoE ϵ4 with cognitive decline was not limited to memory-related functions, as has been suggested by Bondi et al.¹⁷

There are several methodological issues that might have affected our findings. First, ceiling or floor effects on the tests might have influenced our ability to detect changes in the different diagnostic groups. This is less likely to be a problem with the 107-point CAMCOG used as an outcome than with a shorter mental status test. The CAMCOG has a high ceiling, with very few people scoring 100%. In addition, the scores are well distributed, even among individuals with dementia (range in score, 29-77).

A second issue is related to the characteristics of the subjects who dropped out over the 4-year period. Dropout would have altered the results if it had affected the proportion of apoE ϵ4 carriers, which was not the case. The proportions of ϵ4 allele carriers at baseline and at the end of the study were not significantly different. However, not all individuals with the ϵ4 allele will develop dementia or decline cognitively. If there was a disproportionate dropout of nonsusceptible ϵ4 carriers, which seems unlikely, it would explain the differences in the group with normal cognitive function. Third, the small sample size in the minimal dementia and dementia groups might have affected the results. However, combining the 2 groups did not change the conclusions. In addition, including the patients with incident AD did not change the conclusions. Finally, it is possible that the findings in the normal cognitive function group mainly reflected the decline of individuals who were in the very early stages of dementia. Although it is difficult to completely identify presymptomatic individuals, we did remove those with minimal dementia at baseline and those who subsequently developed minimal or clinical dementia during follow-up.

When the patients with incident AD were removed from the normal cognitive function group, the mean decline in CAMCOG score was lower, but the difference between the carriers and noncarriers of ϵ4 remained significant (P<.001). However, we could not account for incident cases among those who dropped out before the end of the follow-up period.

To our knowledge, this is the first study that compares the rate of cognitive decline associated with ϵ4 status across diagnostic groups of individuals selected from the same source population. This comparison removes some of the barriers to interpreting the results of various studies based on patient groups identified in different ways, at different stages of dementia, and whose progression is evaluated with different instruments. This study has the advantage of including subjects identified in a community-based survey and providing an internally valid comparison of groups at different stages of dementia. Furthermore, we were able to collect data multiple times over a 4-year period, thus allowing us to estimate more reliably the rates of cognitive decline in the sample.

Our findings of the risk for incident dementia (mainly AD) associated with ϵ4 are consistent with those reported by Petersen and colleagues⁷ in a sample of individuals with memory impairment and by Evans et al²⁷ in a sample of community residents. Our finding of a lack of association between apoE ϵ4 and cognitive decline in symptomatic subjects is consistent with findings from recent clinical studies of patients with AD.^10-16

In summary, the results of this study suggest that once cognitive symptoms are manifest, ϵ4 allele dose cannot predict an expected rate of cognitive decline, as some authors suggest.^28-30 Thus, the role of apoE ϵ4 in the development of AD might be less direct than has been suggested.^1,31-33

Views on the role of ϵ4 in the cascade of neuropathological processes in AD, in particular around the time of clinical onset, may have to be reconsidered in light of findings from this and previous studies. Our study suggests that the pathophysiological mechanism of apoE ϵ4 may more strongly affect the timing of the onset of the disease than the progression of the disease once the clinical threshold is reached. This is consistent with Roses'³⁴ suggestion that the primary disease process becomes autocatalytic when symptoms first appear. Another possibility suggested by Plassman and Breitner³⁵ is that massive loss of primary targets of the disease process induces secondary degeneration of neural structures. Such degeneration might produce rapid decline in cognitive performance, not necessarily influenced by apoE ϵ4. Finally, this study supports the suggestion that at present, apoE ϵ4 genotyping has few practical consequences for the diagnostic processes, management, and care of symptomatic patients.³⁶

Accepted for publication December 21, 1997.

The Amsterdam Study of the Elderly (AMSTEL) is supported by grants from the Netherlands Health Research Promotion Programme, the Hague, and the National Fund of Mental Health, Utrecht, the Netherlands.

We are grateful to Piet Eikelenboom, MD, PhD, for his comments on an earlier draft of this article.

Reprints: Cees Jonker, MD, PhD, AMSTEL-project, Faculteit SCW, Vrije Universiteit, De Boelelaan 1081, 1081 HV Amsterdam, the Netherlands (e-mail: C.Jonker@scw.vu.nl).