Abstract

BACKGROUND

In numerous studies of older adults in the United States, blacks have performed worse, on average, than whites on tests of cognitive function.^1,2 Furthermore, many studies have documented higher prevalence and incidence of dementia, including Alzheimer disease dementia (AD) among blacks than among whites.³ Dementia is typically the consequence of decline in cognitive function, thus racial differences in cognition and dementia risk would be expected to be mirrored in faster longitudinal cognitive decline among blacks than among whites. Investigations of this question, however, have generated mixed results.^1,2,4–9 Although many of these studies recruited community participants, and several enrolled thousands of participants, most included few black participants, making it challenging to compare in detail the cognitive aging experiences of black and white older adults.

The population-based Chicago Health and Aging Project (CHAP) evaluated more than 10,000 older adults, 60% of whom were black, over a period spanning nearly two decades. We used data from CHAP to estimate racial differences in cognitive level and rate of cognitive decline, and to update previous estimates of racial differences in dementia risk. Compared with data used for these previous estimates, the updated data comprise over three times as many dementia assessments and dementia cases.¹⁰ We also compared black and white participants’ cognitive outcomes according to whether they carried an APOE ε4 allele, a major genetic risk factor for AD. Finally, we quantified the extent to which racial differences in the outcomes were mediated by educational attainment.

METHODS

RESULTS

White participants were, on average, 3 years older than black participants; the groups contained similar percentages of men (Table 1). In unadjusted analyses, black participants’ baseline global cognitive scores were substantially lower than whites’ scores; rates of change were not substantially different. Of the 2,144 participants selected for clinical evaluation and previously determined to be dementia-free, 474 developed dementia (22%); 441 were identified as having AD, of whom 275 were black.

Table 1

Characteristics of participants, by race, in the analyses of cognitive level, cognitive decline, and incident dementia.

	Black participants	White participants	Difference,ref: white (95% CIa)
Analyses of cognitive level and cognitive decline
	N=4968	N=2767
Mean age, years, at baseline (SDa)	71 (6)	74 (7)	−3.0 (−3.3, −2.7)
Male (%)	37	38	−1 (−3, 2)
Mean years of formal education (SD)	12 (3)	14 (3)	−2.4 (−2.6, −2.3)
Composite childhood socioeconomic resources score (SD)b	−0.1 (0.7)	0.3 (0.7)	−0.44 (−0.47, −0.41)
APOE ε4 allele carrierc (%)	37	26	11 (8, 14)
Mean baseline global cognitive score (SD)	−0.2 (1.0)	0.4 (0.9)	−0.59 (−0.63, −0.55)
Mean 5-year change in global cognitive score (SD)	−0.25 (0.63)	−0.26 (0.68)	0.01 (−0.02, 0.04)
Analyses of incident dementia
	N=1170	N=974
Mean age, years, at baseline (SD)	72 (5)	75 (6)	−3.0 (−3.6, −2.6)
Male (%)	37	39	−2 (−6, 2)
Mean years of formal education (SD)	12 (3)	14 (3)	−2.4 (−2.6, −2.1)
Composite childhood socioeconomic resources score (SD)b	−0.1 (1.1)	0.5 (1.1)	−0.5 (−0.6, −0.4)
APOE ε4 allele carrierc (%)	35	26	9 (5, 13)
Incident ADa case (%)d	24	17	6 (3, 10)
Incident dementia case (%)d	25	18	7 (3, 10)

Open in a separate window

^aCI: confidence interval. SD: standard deviation. AD: Alzheimer disease dementia.

^bHigher values correspond to more socioeconomic resources and less socioeconomic disadvantage during childhood. Data available for 4587 black and 2748 white participants in the analyses of cognitive level and decline, and for 1107 black and 971 white participants in the analyses of incident dementia.

^cFrom a subsample of participants: 2997 black and 1257 white participants (cognitive decline); and 1105 black and 843 white participants (dementia).

^dOver the entire course of follow-up.

Mediation by education

On average, black participants had two fewer years of formal education than did whites (Table 1). In age-, sex-, and race-adjusted analyses, more years of education corresponded to higher baseline global cognitive score (0.40 standard unit per each additional 4 years; 95% CI, 0.37 to 0.42) and lower AD risk (RR per each additional 4 years, 0.74; 95% CI, 0.68 to 0.80). We did not further consider cognitive trajectory in these analyses as cognitive decline was not notably faster among blacks in our data.

Educational attainment, as measured by years of education, appeared to mediate a substantial fraction but not the totality of the racial differences in baseline cognitive score and AD risk (Table 5). Under the hypothetical scenario in which education was “controlled” such that each black participant’s educational level took on the level it would have been had the participant been white, all covariates being equal, black participants’ baseline global cognitive scores were an average of 0.45 standard units lower than whites’ scores (95% CI, −0.49 to −0.41), a difference smaller than without controlling years of education (−0.69; Table 5), and translating to about 35% of the total effect of race on cognitive performance mediated through years of education. The racial disparity in cognitive performance was narrower among those with more years of education (P_interaction < 0.001), echoing previous analyses.^26,30 This interaction manifested itself in the controlled direct effects. In the scenario in which everyone obtained 12 years of education, the black–white difference in global cognitive score dropped to 0.57 standard units (95% CI, −0.62 to −0.53), but in the scenario in which everyone obtained 16 years of education, the difference dropped even further to −0.29 standard units (95% CI, −0.35 to −0.24), eliminating 58% of the total racial disparity in performance. There was no notable interaction between race and education in relation to AD risk. Thus, equally distributing of years of education across race, regardless of the educational level, would eliminate about 37% of the excess relative risk among black participants. (For more detail, see eAppendix [Supplemental text; Mediation by education and eTable 1a and eTable1b].)

Table 5

Mediation by education of the associations of race with cognitive score and Alzheimer’s dementia risk.

	Adjusteda difference (95% CI) between black and white participants’ baseline cognitive scores	Adjusteda relative risk (95% CI) of incident Alzheimer’s dementia among black versus white participants
	N = 7335	N = 2827
Estimated total effect	Estimated effect of race (black compared with white) on cognitive performance through all causal pathways.
	−0.69 (−0.73, −0.65)	1.49 (1.05, 2.01)
Estimated natural direct effect	Estimated effect of race (black compared with white) on cognitive performance through pathways other than those involving years of education.
	−0.45 (−0.49, −0.41)	1.30 (0.91, 1.78)
Estimated controlled direct effect	Estimated effect of race (black compared with white) on cognitive performance under the scenario in which everyone attains the same specified years of education.b
Everyone attains 12 years of education	−0.57 (−0.62, −0.53)	1.30 (0.91, 1.78)
Everyone attains 16 years of education	−0.29 (−0.35, −0.24)
Proportion of effect eliminated by setting years of education to specified level	Proportion of difference eliminatedc	Proportion of excess relative risk eliminatedb,d
Everyone attains 12 years of education	17%	37%
Everyone attains 16 years of education	58%

Open in a separate window

^aAdjusted for age, sex, and childhood socioeconomic resources.

^bWith no interaction between race and education with respect to Alzheimer’s dementia risk, the estimated natural direct effect and controlled direct effects are equivalent, and the controlled direct effect and proportion of effect eliminated do not vary by years of education.

^c[(Total Effect - Controlled Direct Effect)/(Total Effect)].

^d[RR(Total Effect) - RR(Controlled Direct Effect)]/[RR(Total Effect)-1]. Excess relative risk is the portion of the relative risk that exceeds 1, i.e., RR − 1.

DISCUSSION

In this large, population-based study of older adults, black participants performed worse, on average, on cognitive tests and were more likely to have prevalent AD than white participants at baseline; they also were at higher risk for incident AD. By contrast, the rate of cognitive decline among blacks differed little from or was slightly slower than the rate among whites. Collectively, these results suggest that persistent racial differences in cognitive level—rather than differences in cognitive decline in old age—likely underlie the higher risk of incident AD among blacks.

The large deficits among blacks in average level of cognitive test performance fall in line with the vast body of previous research. Moreover, the elevated risk of incident AD and all-cause dementia among blacks in CHAP is consistent with the age- and education-adjusted AD hazard ratio (HR) of 2.6 (95% CI, 1.6–4.2) from the Washington Heights-Inwood Community Aging Project³¹ and the age-, sex-, and APOE ε4-adjusted all-cause dementia HR of 1.44 (95% CI, 1.20–1.74) from the Dynamics of Health, Aging, and Body Composition study.³² By contrast, in the Aging, Demographic, and Memory Study, blacks and whites did not differ substantially in incident AD risk, after adjustment for age, education, sex, and APOE ε4 (cumulative incidence ratio [CIR]=1.22; 95% CI, 0.58–2.57), and risks of incident all-cause dementia in the two groups were nearly identical (CIR=0.98; 95% CI, 0.50–1.95).³³ An older study from the Duke Established Populations for Epidemiologic Studies of the Elderly reported a small, imprecisely measured difference between blacks and whites in three-year dementia incidence; the analyses were not age-adjusted, however, and included only 77 dementia cases.³⁴ In the Cardiovascular Health Study, compared with white participants, the age-adjusted incidence rate of AD among blacks was 81% higher, and the corresponding rate of all-cause dementia was 72% higher. However, the study’s diagnostic procedures varied by race, and the investigators posited that this could largely explain the racial difference in incidence.³⁵

Previous studies of cognitive decline, generally with smaller sample sizes, have found either little racial difference or slower decline among blacks.^1,2,4,5,9 Ours is the third study to observe substantially slower decline among blacks on tests of executive function.^1,9 Of the two previous studies reporting faster cognitive decline among blacks,^6,7 one adjusted for baseline cognitive score, an approach that can severely bias results in the direction of faster decline among blacks when blacks perform markedly worse at baseline.³⁶

The racial disparity in dementia risk in CHAP was substantial among persons not carrying an APOE ε4 allele, but absent among carriers. Equivalently, whereas ε4 carriership corresponded to a doubling of risk among white participants, it corresponded to no increase in risk among blacks. This pattern held on the additive (risk difference) scale as well. Similar results emerged from previous analyses of data from CHAP¹⁰ and two other cohorts.^37,38 Nonetheless, the ε4 allele was associated with increased risk (P ≤ 5 × 10⁻⁸) in a genome-wide association study of data from 5896 black participants of 15 studies.³⁹ The limited evidence on cognition and cognitive decline has been mixed as well. In CHAP and two other studies,^7,40 racial differences in cognitive level did not further depend on ε4 carriership, as opposed to another study in which ε4 carriership was associated with worse memory performance in white but not black participants.³⁸ Likewise, in data from CHAP and another cohort,⁷ ε4 carriership had no bearing on racial differences in cognitive decline; in two other cohorts, blacks’ decline in semantic memory and working memory scores was slower than whites’ cognitive decline among ε4-positive persons, whereas no racial differences in cognitive decline were evident among ε4-negative persons.⁴¹

On their surface, the findings on cognitive decline contradict the findings on cognition and dementia risk. Accurately characterizing and effectively intervening on racial disparities in cognitive dysfunction and dementia hinges on understanding these findings. The findings could reflect differences in cognitive reserve caused by racial disparities over the lifespan in, for example, education (both in amount and quality), access to material and social resources, exposure to discrimination, and exposures to neurotoxicants.^27,42 With higher cognitive reserve by older adulthood, on average, white individuals have “farther to fall” cognitively than black individuals before reaching the functional threshold of clinical dementia, so that even if both groups have the same rate of cognitive decline, blacks have poorer cognitive function and disproportionately develop dementia. If this explanation is correct, then intervening on the promoters of cognitive reserve among blacks should enhance cognitive functioning and reduce dementia risk in this group. Another explanation for the discrepancy between the cognitive decline and other results posits that the cognitive and dementia findings—particularly the cross-sectional differences in cognitive performance—reflect measurement bias.^3,43–45 Race-dependent misclassification of dementia status would clearly be problematic and call for unbiased approaches to assessment. Notably, some cognitive tests do not appear to be susceptible to marked, racially differential measurement error.^46,47

Years of formal education accounted for much but not all of the observed racial disparities in global cognitive score and AD risk. The counterfactual approach we used offers several advantages over the conventional approach of adjusting for putative mediators. However, the education measure—years of education—holds meaning that depends on race, particularly in this generation of participants, representing different drivers, quality, acquired skills,⁵ and socioeconomic⁴⁸ and health³⁰ consequences. Thus, our analyses likely failed to fully level the educational playing field, possibly underestimating the extent to which education mediates these associations. For example, in the Washington Heights–Inwood Community Aging Project indicators of educational quality explained racial differences in performance beyond what was explained by years of education.²⁶ Nonetheless, our findings, together with others, provide support for the claim that early-life education, which social policy can influence at the broad population level, may be one strategy for reducing racial disparities in late-life cognitive health.

Our study included more black participants—often by an order of magnitude—than any previous comparisons of cognitive decline and dementia risk in blacks and whites. CHAP also systematically drew its participants from the population, permitting a potentially more accurate characterization of the cognitive aging experience among community-dwelling black and white older adults. Nonetheless, several limitations and challenges to the interpretation of our results warrant mention.

First, our findings—and those of previous studies—may have been influenced by differential selection. Enrollment is less likely and post-enrollment attrition more likely among individuals with poor cognition.²⁹ If race is also related to enrollment and attrition, bias in estimated racial differences in cognition, cognitive decline, and dementia risk could result. Findings from inverse probability-of-continuation-weighted analyses indicate that differential post-enrollment attrition, including post-enrollment mortality, does not explain our cognitive decline findings.

However, racial disparities in mortality may shape the initial composition of an older adult cohort, because, among the birth cohorts that generated older adult study populations, blacks were much less likely than whites to survive to the age of 65, a typical minimum age of enrollment (e.g., among those born in 1939–1941, 38% of blacks versus 63% of whites survived to age 65⁴⁹). Specifically, blacks who survived to enrollment age may have been more likely than surviving whites to possess characteristics that both promoted their survival and conferred cognitive benefits.⁵⁰ This type of selection is more challenging to address, and if it biases racial comparisons of cognitive outcomes in older age at all, it likely would result in an underestimate of blacks’ relative risk of dementia and deficit in cognitive performance. It could also generate the observation of slower cognitive decline among blacks in the true absence of any racial difference, or even in the true existence of faster decline among blacks.⁵¹ All of these biases can be exaggerated in progressively older strata, which are progressively more selected; e.g., as observed in CHAP and in another cohort,⁹ the pattern of slower decline in global and executive function among blacks was more pronounced with older age. Without “correcting” this bias, if it exists, our study’s estimates describe the experiences of individuals in the population who have survived to an older age; they are essentially a demographic record of a specific time. But the comparisons might not fully capture the total causal effect—e.g., the combined effect of social disadvantage and socioeconomic inequality—of being black versus being white on cognitive aging.^52,53 This distinction is important: underestimated racial differences in cognitive aging imply a smaller impact of intervening on the causes of racial disparity in health.⁵⁴ Because of this uncertainty, it would be incorrect to conclude, for example, that interventions to slow cognitive decline should be targeted to white individuals.

Selection bias also has been invoked to explain the apparently diminished APOE ε4 effect on dementia risk among black Americans.^38,55 Alternatively, blacks’ dementia risk is largely a function of cognitive level rather than decline, and because the ε4 allele is more strongly related to cognitive decline than level, it may wield little influence on blacks’ dementia risk. Others have proposed that the high prevalence of other unmeasured risk factors among blacks could effectively crowd out ε4’s relative contribution to risk, or, equivalently, that the overwhelming dominance of ε4 as a risk factor diminishes the relative contribution of race to risk.³⁸ This purely mathematical construction holds because these two factors—the ε4 allele and being a black American—affect risk, not because of their lack of an effect,⁵⁶ a critical point in translating these results to public health and clinical practice.

In conclusion, these findings, from one of the largest studies to date of older black and white Americans, sharpen a pattern of results from other cohorts and suggest that the higher risk of incident AD among blacks observed in this and other studies may reflect persistent differences in cognitive level rather than differences in cognitive decline in older age. Future research is warranted to carefully characterize the causal ties between the manifestations of race and cognitive aging.

Supplementary Material

Acknowledgments

Footnotes

References