Associations of APOE ε4 With Health and Financial Literacy Among Community-Based Older Adults Without Dementia

Abstract

Objectives

Older adults often exhibit low health and financial literacy, but the reasons why remain unclear. One possibility is that those older adults at high risk for developing dementia demonstrate low literacy even in the absence of marked cognitive impairment. We therefore examined associations of health and financial literacy with the APOE ε4 allele, the chief genetic risk factor for Alzheimer’s disease, among older adults without dementia.

Method

Participants were 487 older adults without dementia enrolled in the Rush Memory and Aging Project (mean age = 83, mean years of education = 15, 77% female, 91% non-Hispanic White). Participants underwent APOE genotyping and assessments of cognition, health literacy, and financial literacy. Health and financial literacy scores were also averaged into a total literacy score.

Results

ε4 was associated with lower total and health literacy, with a trend toward an association with lower financial literacy, after adjustment for age, sex, and education. Associations of ε4 with lower total and health literacy persisted after further adjustment for global cognitive function and 5 specific cognitive domains.

Discussion

ε4 affects literacy even in the absence of clinical dementia and does so relatively independent of performance on traditional cognitive tests.

Although the term “literacy” is often used to refer to the ability to read and write, more recent conceptualizations refer instead to a dynamic, lifelong learning process that involves the acquisition and application of relevant knowledge—for example, about health and finances—to circumstances encountered in everyday life (Baker, 2006). Elaborating on this, per the United Nations Educational, Scientific, and Cultural Organization (UNESCO), literacy involves “the ability to identify, understand, interpret, create, communicate and compute, using printed and written materials associated with varying contexts. Literacy involves a continuum of learning in enabling individuals to achieve their goals, to develop their knowledge and potential, and to participate fully in their community and wider society” (UNESCO Education Sector, 2004).

Older adults often struggle in two specific domains of literacy: health and financial literacy (Kobayashi, Wardle, Wolf, & von Wagner, 2016; Lusardi & Mitchell, 2006; Williams et al., 1995). Low health and financial literacy can undermine the many influential health care and financial decisions older adults often face (e.g., selecting optimal health care and prescription drug benefit plans, weighing the pros and cons of end-of-life care options, delineating intergenerational transfers of wealth, budgeting retirement savings) (James, Boyle, J. S. Bennett, & D. A. Bennett, 2012; Lusardi & Mitchell, 2007a, 2007b). Moreover, low health and financial literacy have been linked to various negative health outcomes among older adults, including lower participation in health-promoting behaviors (J. S. Bennett, Boyle, James, & D. A. Bennett, 2012; Scott, Gazmararian, Williams, & Baker, 2002), poorer health status (Baker, Parker, Williams, Clark, & Nurss, 1997; Sudore, Mehta, et al., 2006; Wolf, Gazmararian, & Baker, 2005), and higher rates of mortality (Baker et al., 2007; Baker, Wolf, Feinglass, & Thompson, 2008; Sudore, Yaffe, et al., 2006).

Most studies examining why older adults often exhibit low health and financial literacy have focused on cohort effects, such as exposure to financial and health education in high school and at work (Gazmararian et al., 1999; Lusardi & Mitchell, 2007a, 2007b). However, emerging data suggest that lower literacy in old age is related to the development of dementia (Wilson, Yu, James, Bennett, & Boyle, in press); thus, traditional risk factors for dementia may be associated with lower literacy even in the absence of marked cognitive impairment. The apolipoprotein E (APOE) ε4 allele is particularly well suited to test this idea because it is the chief genetic risk factor of clinical Alzheimer’s dementia (Corder et al., 1993; Raber, Huang, & Ashford, 2004) and is not intertwined with sociocultural and other risk factors for dementia (e.g., education, income), many of which also affect literacy or could have bidirectional relationships with literacy (Chin, Negash, & Hamilton, 2011; Gazmararian et al., 1999; Sisco et al., 2015; Yaffe et al., 2013). In addition, unlike many other risk factors for dementia, the neuropathological underpinnings by which APOE ε4 leads to the clinical manifestations of dementia are well understood, as ε4 has been shown to work primarily through pathologic hallmarks of Alzheimer’s disease (i.e., beta amyloid plaques and neurofibrillary tangles) (Bennett et al., 2003; Yu, Boyle, Leurgans, Schneider, & Bennett, 2014).

While no study, to our knowledge, has examined the relation of ε4 status to health and financial literacy, there is reason to believe that literacy might be sensitive to the early effects of ε4 among older adults without dementia. Recent evidence suggests that patients with mild cognitive impairment (MCI) exhibit deficits in higher-order language functions, including semantic integration during sentence processing (Payne & Stine-Morrow, 2014). Because literacy depends in part on semantic integration for the acquisition of general and domain-specific knowledge, literacy might also decline prior to the onset of dementia. Moreover, literacy is highly complex, involving not just semantic integration of information into general and domain-specific knowledge bases but also computational abilities (Chin et al., 2011, 2015). Given its multifaceted cognitive underpinnings, literacy is thought to rely on the coordinated activity of a widely distributed neural network (Han et al., 2014), and this might make literacy particularly vulnerable to early ε4-related neuropathology.

The current study tested the hypothesis that the presence of one or more ε4 alleles is associated with lower health and financial literacy in a cohort of more than 450 older adults without dementia. In primary analyses, we examined associations of ε4 with literacy measures after adjusting for age, sex, and education and then after further adjusting for global cognition. In secondary analyses, we examined ε4-literacy associations after adjusting for age, sex, education, and one of five specific cognitive domains: episodic memory, semantic memory, working memory, perceptual speed, or visuospatial abilities. These secondary analyses were conducted because ε4 has been shown to preferentially affect episodic memory in old age (Caselli et al., 2004, 2009; Wilson et al., 2002b).

Method

Participants

Participants were older adults without dementia from the Rush Memory and Aging Project, an ongoing longitudinal clinicopathological study of common chronic conditions of aging (Bennett, Schneider, Buchman, et al., 2012). Participants were recruited from about 40 senior housing facilities and retirement communities in the greater Chicagoland area, as well as through subsidized housing, social service agencies, and church groups to obtain a range of socioeconomic status. Participation in the study includes risk factor assessment and detailed annual clinical evaluations. The study was approved by the Institutional Review Board of Rush University Medical Center. Informed consent was obtained from each participant following a comprehensive review of the risks and benefits of study participation.

The Rush Memory and Aging Project started in 1997, and the assessments of health and financial literacy were added to the parent study in 2010. At the time of the current analyses, 1,769 participants had completed the baseline evaluation for the parent study; of those, 590 died before the literacy assessment, 81 withdrew before the literacy assessment, and 80 were deemed ineligible for the literacy assessment due to severe difficulties with language, hearing, vision, or understanding or due to having moved out of the geographical area. Of the remaining 1,018 potentially eligible participants, 43 refused the literacy assessment, 71 had yet to complete the literacy assessment, and 904 participants had completed the literacy assessment. Of the 904 participants, 53 had dementia and were excluded, 330 had genotype data pending, and 26 had missing literacy data. Eight participants with the ε2/4 genotype were excluded because ε2 is protective against cognitive decline (Corder et al., 1994). This left 487 participants eligible for analyses, of which 100 (21%) were ε4 carriers (93 with ε3/4; 7 with ε4/4) and 387 were noncarriers (2 with ε2/2; 68 with ε2/3; 317 with ε3/3). Analyses comparing the demographics of the study sample with those who refused or had yet to complete the literacy assessment revealed that two groups did not differ with respect to age, sex, or ε4 allele status (all ps > .05), while those who refused or had yet to complete the literacy assessment had a lower level of education (p = .017) and were more likely to be a racial minority (p < .001).

Clinical Diagnosis

All participants underwent uniform structured clinical evaluations, as previously described (Bennett et al., 2005; Bennett, Schneider, Buchman, et al., 2012). The diagnosis of dementia was made by clinicians experienced in the assessment of older adults and followed the National Institute of Neurologic and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association criteria. These criteria require a history of cognitive decline and evidence of impairment in at least two cognitive domains (McKhann et al., 1984). All participants in the current analyses were without dementia at the time of the literacy assessment. Clinical evaluations were conducted by clinicians blinded to participants’ APOE genotype and performance on the literacy assessment.

Assessment of Literacy

Literacy was measured via 32 items requiring general knowledge and domain-specific knowledge of health and financial concepts, in addition to computational abilities, including numeracy (i.e., the performance of simple calculations) (Bennett, Boyle, et al., 2012; James et al., 2012). Twenty-three items addressed financial literacy, many of which were modeled off the Health and Retirement Survey (Lusardi & Mitchell, 2007a, 2007b). These items required numeracy and/or knowledge of finances (e.g., stocks, bonds, compound interest) (the following are examples: Question: If a television set is on sale for $899, which is $200 off its normal price, what is the normal price? (1) $699, (2) $1,099, (3) $1,299; Question: A mutual fund is an investment that holds what? (1) Only stocks, (2) Only bonds, or (3) Stocks and bonds). Nine items assessed health literacy. These items addressed topics including Medicare and Medicare Part D, following prescription instructions, comprehension of drug risk, and common causes of death in older adulthood (the following are examples: Question: Medicare Part D covers which of the following? (1) Inpatient hospital services, (2) Outpatient physician services, (3) Prescription drug benefits; Question: True or false? Medicare routinely covers costs associated with extended long-term care, such as nursing home care lasting more than 1 month. For the complete health and financial literacy assessments, please see James et al. (2012). Item format was either multiple choice or true/false, and each item was scored as correct or incorrect. Health and financial literacy scores were expressed as the percent correct out of total items (from 0% to 100%). Total literacy was the average of these two percentages. Cronbach’s coefficient alphas for total and financial literacy were .763 and .736, respectively, suggesting adequate internal consistency. Cronbach’s coefficient alpha was lower for health literacy (.472); this likely reflects the brevity of the measure (nine items) and the fact that it assesses varied aspects of health and health care knowledge and the application of this knowledge to different situations that older adults commonly face. Notably, our measures of total, health, and financial literacy show robust associations with multiple relevant outcomes, including better health care and financial decision making (James et al., 2012), better daily functioning and mental health, and more frequent participation in cognitive, physical, and social activities (Bennett, Boyle, et al., 2012).

APOE Genotyping

Blood was collected at study sites via BD Vacutainer CPT cell preparation tubes containing sodium citrate and kept at room temperature. The mononuclear cell layer was segregated within 24h of blood collection. DNA was extracted from approximately 2–3 million mononuclear cells via Flexigene DNA extraction kits (Qiagen, Valencia, CA) and quantified in 96-well plates using Quant-iT PicoGreen dsDNA detection assay kits (Molecular Probes, Eugene, OR). DNA was typically extracted from blood but was sometimes from frozen postmortem cerebellar tissue. Genotyping was conducted by investigators blinded to study data using high-throughput sequencing of codon 112 (position 3,937) and codon 158 (position 4,075) of exon 4 of the APOE gene on chromosome 19 (Boyle, Buchman, Wilson, Kelly, & Bennett, 2010).

Assessment of Cognition

Cognition was measured via a battery of 21 individual performance tests, as previously described (Bennett et al., 2005). The battery included the Mini-Mental State Examination, but these scores were used only for descriptive purposes and not used in study analyses. Another test, Complex Ideational Material, was used only for diagnostic classification purposes and was not included in the cognitive domains because its distribution was highly negatively skewed (Wilson et al., 2002a). Scores on the remaining 19 tests were used to create composite indices of global cognitive function and five specific cognitive domains, as we have done in previous publications (for reviews of these studies, see Bennett, Schneider, Buchman, et al., 2012 and Bennett, Schneider, Arvanitakis, et al., 2012). The five specific cognitive domains were as follows: (1) episodic memory (Word List Memory, Word List Recall and Word List Recognition from the procedures established by the CERAD; immediate and delayed recall of Logical Memory Story A and the East Boston Story), (2) semantic memory (Verbal [Category] Fluency, Boston Naming, and the National Adult Reading Test), (3) working memory (Digit Span subtests [forward and backward] of the Wechsler Memory Scale-Revised and Digit Ordering), (4) perceptual speed (oral version of the Symbol Digit Modalities Test, Number Comparison, Stroop Color Naming, and Stroop Word Reading), and (5) visuospatial abilities (Judgment of Line Orientation and Standard Progressive Matrices). To calculate composite scores, raw scores on the 19 individual tests were converted to z scores using the baseline mean and SD of the entire Rush Memory and Aging Project cohort, of which the current sample is a subset. To compute global cognitive function, the z scores of all 19 individual tests were averaged. To compute composite indices of the specific cognitive domains, the z scores of individual tests from their respective cognitive domain were averaged. The grouping of individual tests into the five specific cognitive domains was both conceptually and empirically driven; the groupings are supported by factor analysis using data from different cohorts and also by a high degree of agreement between our conceptually based groupings and those derived from factor analysis (Rand’s measure = .79, p < .01) (see Wilson et al., 2002a). The specific cognitive domains also demonstrate good internal consistency (Cronbach’s coefficient alphas range from .701 to .765), as does our measure of global cognition (Cronbach’s coefficient alpha = .783).

Demographics

Age was calculated from date of birth, based on self-report at the initial study evaluation, and the date of the literacy assessment. Sex and education (years of schooling) were based on self-report at the initial study evaluation.

Statistical Analysis

We first examined bivariate associations of APOE ε4 allele status and literacy measures (i.e., total, health, and financial) with demographics, global cognitive function, and the five specific cognitive domains via t tests, χ², or Pearson correlations, as appropriate. We next examined associations of ε4 allele status with literacy measures via a series of multivariable linear regression models with literacy measures as the continuous outcome. In primary regression models, we first adjusted for age, sex, and education, and then further adjusted for global cognition. We then conducted secondary regression models, which adjusted for age, sex, and education, and one of each of five specific cognitive domains (in separate models).

Results

Characteristics of the Sample

Mean percent correct on measures of total, health, and financial literacy in the overall sample were 67%, 62%, and 73%, respectively.

Bivariate analyses examining associations of APOE ε4 with demographics and measures of cognition and literacy are reported in Table 1. Age, sex, and race did not differ by ε4 allele status, but education was higher among ε4 carriers compared to noncarriers. Global cognition and the specific cognitive domains did not differ by ε4 allele status. Health literacy was lower among ε4 carriers compared to noncarriers, while total and financial literacy did not differ by ε4 allele status.

Bivariate analyses examining associations of literacy scores with each other, demographics, and measures of cognition are reported in Table 2. Total, health, and financial literacy were positively correlated with one another, education, global cognition, and each of the specific cognitive domains and negatively correlated with age. Independent t tests showed that measures of total and financial literacy were higher in males than females (total literacy: mean for males = 71.7%; mean for females = 65.9%; p < .001; financial literacy: mean for males = 82.8%; mean for females = 69.6%; p < .001), but health literacy did not differ by sex (health literacy: mean for males = 60.6%; mean for females = 62.2%; p = .419).

Associations of APOE ε4 With Literacy

Regression models examining associations of ε4 with total, health, and financial literacy are reported in Table 3. In models adjusted for age, sex, and education, the presence of the ε4 allele was associated with lower total and health literacy, with a trend toward an association with lower financial literacy. To contextualize the magnitude of these findings, ε4-related effects on literacy were translated to age-related effects on literacy. The ε4-related reduction in total literacy was equivalent to the reduction associated with being about 8 years older, and the ε4-related reduction in health literacy was equivalent to the reduction associated with being about 10 years older. In regression models that further adjusted for global cognition, associations of ε4 with lower total and health literacy were attenuated by 31% and 24%, respectively, but remained significant. The association of ε4 with financial literacy was not significant after adjusting for global cognition.

Secondary Analyses

Because ε4 has been shown to preferentially affect particular cognitive domains, specifically episodic memory (Caselli et al., 2004, 2009; Wilson et al., 2002b), we conducted secondary regression models examining associations of ε4 with total, health, and financial literacy after adjusting age, sex, education, and each of the five specific cognitive domains (in separate models). Results resembled those of the models that adjusted for global cognition (Table 3) in that associations of ε4 with lower total and health literacy remained significant, whereas associations of ε4 with financial literacy generally were not significant. Episodic memory had the strongest attenuating effect among the five specific cognitive domains, reducing associations of ε4 with total literacy and health literacy by 23% and 19%, respectively.

Discussion

We examined the relation of the APOE ε4 allele with measures of total, health, and financial literacy in a sample of 487 community-based older adults without dementia. After adjustment for age, sex, and education, the presence of one or more ε4 alleles was associated with lower total and health literacy and trended toward an association with lower financial literacy. Associations of ε4 with lower total and health literacy persisted in regression models that further adjusted for global cognition and five specific cognitive domains (episodic memory, semantic memory, working memory, perceptual speed, and visuospatial abilities). These findings suggest that ε4 affects literacy in the absence of clinical dementia.

The present study is the first to our knowledge to relate a well-established genetic risk factor for Alzheimer’s dementia with literacy in late life. Associations of ε4 with lower literacy are particularly striking because ε4 is only sometimes associated with lower cognition among samples of older adults without dementia (Bretsky, Guralnik, Launer, Albert, & Seeman, 2003; DeCarlo et al., 2015; Finkel, Reynolds, Larsson, Gatz, & Pedersen, 2011; Small, Graves, et al., 2000; Small, Rosnick, Fratiglioni, & Bäckman, 2004; Wang et al., 2015). We suspect that ε4-literacy associations were observed because literacy is complex and multidimensional, requiring the integration of general knowledge, domain-specific knowledge, and computational abilities, and therefore likely relies on the coordinated activity of a widely distributed neural network. Our results suggest that this neural network may be susceptible to early ε4-related neuropathology (i.e., beta amyloid plaques and neurofibrillary tangles). That the association of ε4 with lower literacy occurred relatively independent of cognition, including episodic memory, which is preferentially affected by ε4 in old age (Caselli et al., 2004, 2009; Wilson et al., 2002b) and highly correlated with our literacy measure (J. S. Bennett, Boyle, James, & D. A. Bennett, 2012; Boyle et al., 2013), suggests that ε4 influences literacy at least in part through abilities that are outside the purview of traditional, performance-based neuropsychological measures. While previous studies have examined the relationship of ε4 with noncognitive factors, including responses on standard inventories of temperament/personality and psychiatric symptoms, most have not found an association (Farlow et al., 2004; Jorm et al., 2003; Hillemacher et al., 2006; Montag et al., 2014; Slifer, Martin, Gilbert, Haines, & Pericak-Vance, 2009; Keltikangas-Jarvinen, Raikkonen, & Lehtimaki, 1993; Tsai, Yu, & Hong, 2004), making these factors unlikely to account for the relation of ε4 with lower literacy.

Current conceptualizations of the cognitive underpinnings of literacy, which posit that literacy requires general knowledge, domain-specific knowledge, and computational abilities, align well with our results (e.g., the process-knowledge model of health literacy; Chin et al., 2011, 2015; also see Salthouse (2012) regarding financial literacy). Our cognitive measures were strongly correlated with literacy, including semantic memory and working memory, which correspond most closely to general knowledge and computational abilities, respectively. These conceptualizations also might help explain why associations of ε4 with health literacy persisted after adjusting for cognition, whereas those with financial literacy generally did not. Such a pattern might suggest that health literacy relies particularly heavily on domain-specific knowledge about health and health care (which was not assessed via our cognitive battery). By contrast, financial literacy might rely particularly heavily on computational abilities, consistent with the prominent role of numeracy in financial literacy. That working memory, which is closely linked to numeracy, was more strongly associated with financial literacy than health literacy (r = .41 and r = .28, respectively) provides some support for this.

Although the basis of ε4-literacy associations remains unclear, one possibility is that ε4-related changes in goal-orientedness and other frontal lobe-mediated abilities may play a role. We suggest this because literacy requires not just those abilities assessed by traditional cognitive tests (e.g., general knowledge, computational abilities) but also domain-specific knowledge (Chin et al., 2011, 2015; Salthouse, 2012). In order to acquire domain-specific knowledge, one must typically engage in various learning opportunities (e.g., attendance at seminars, consultation with experts, self-guided research), and this depends at least in part on goal-oriented abilities, such as self-regulation, rational personal and social decision making, and future orientation, which are mediated by frontal lobe, particularly the ventromedial frontal cortex (Allard & Kensinger, 2014; Damasio, 1996; Malloy & Grace, 2005). Notably, ε4 has been shown to affect the frontal lobe and associated goal-oriented abilities among older adults without dementia. Fludeoxyglucose PET, Pittsburgh Compound B PET, and functional MRI have each revealed ε4-related changes in the ventromedial frontal cortex among cognitively healthy middle- and older-aged adults (Filbey, Chen, Sunderland, & Cohen, 2010; Reiman et al., 1996, 2009), and a recent study demonstrated that ε4 carriers with MCI exhibit greater dysfunction in frontal-mediated, goal-oriented abilities compared to noncarriers with MCI, even though the two groups performed comparably on cognitive tests of executive functions (Mikos, Piryatinsky, Tremont, & Malloy, 2013). This latter finding converges with a large literature demonstrating that performance on traditional cognitive tests is often insensitive to impairment in frontal-mediated, goal-oriented abilities (Cato, Delis, Abildskov, & Bigler, 2004; Damasio, 1996; Shamay-Tsoory, Tibi-Elhanany, & Aharon-Peretz, 2006) and fits nicely with our observation that ε4-literacy associations were often independent of cognition. It also is interesting to note that, while associations of ε4 with financial literacy were less robust than those with health literacy, our group recently identified functional connectivity between the ventromedial frontal cortex and the posterior cingulate cortex as an important neural substrate of financial literacy among older adults without dementia (Han et al., 2014). While additional research is clearly needed, it may be that ε4-related degradation of frontal-mediated, goal-oriented abilities leads to suboptimal engagement in relevant learning opportunities, which in turn, might lower domain-specific knowledge that is crucial to literacy.

The present study has strengths and limitations. Strengths include the detailed assessment of domain-specific health and financial literacy. These same literacy measures are associated with relevant outcomes in older adults without dementia, including financial and health care decision making (James et al., 2012), participation in health-promoting behaviors, and health status (Bennett, Boyle, et al., 2012). A second strength is this study’s large, well-characterized sample of community-based older adults without dementia. Among the many advantages this affords, it allowed us to adjust for global cognition and specific cognitive domains to determine that associations of total and health literacy occurred independent of cognition. Limitations are that the study sample was largely non-Hispanic White and relatively highly educated. This limits generalizability to the overall older adult population. On the other hand, the fact that our results were observed in a relatively restricted sample might speak to the robustness of the association of ε4 with lower literacy. A second limitation is that the internal consistency of our health literacy measure was suboptimal, likely owing to the brevity and broadness of the measure (i.e., the fact that different items tap into very different aspects of health literacy). A third limitation is this study’s cross-sectional design, which precluded us from determining whether ε4 carriers experience intraindividual declines in literacy in old age or, alternatively, whether the observed ε4-literacy associations are long-standing. Longitudinal data collection of literacy measures is ongoing and will be examined as data accrue. Future research should investigate the potential role of goal-oriented, frontal-mediated abilities in explaining ε4-literacy associations, in addition to the neural substrates of ε4-literacy associations, for example, via functional connectivity. Investigation of sociocultural and other established risk factors for dementia (e.g., education, income) in relation to literacy in old age is also warranted. This research has the potential to advance our understanding of how dementia risk factors affect abilities other than those measured by common neuropsychological tests and might inform the design of interventions aiming to improve health and financial literacy, and thereby bolster well-being, among individuals at high risk of developing dementia.

Funding

The study was supported by the National Institute on Aging at the National Institutes of Health (R01 AG17917 and R21 AG30765 to D. A. Bennett, R01 AG34374 and R01 AG33678 to P. A. Boyle, K23 AG40625 to S. D. Han).

Acknowledgments

The authors thank the numerous Illinois residents for participating in the Rush Memory and Aging Project. The authors also thank the staff of the Rush Memory and Aging Project and the Rush Alzheimer’s Disease Center.

Conflict of Interest

None of the authors have any financial conflicts of interest to disclose in relation to this article.

References