A lifespan perspective on cognitive reserve and risk for dementia

Abstract

INTRODUCTION

We addressed whether higher education plays a causal role in reducing dementia risk by comparing two indices of cognitive reserve: education and young adult general cognitive ability (GCA).

METHODS

We conducted a 52‐year survival analysis to examine associations of GCA and education with dementia in 16,619 male conscripts identified in Swedish national health registries born between 1936 and 1958 and with available data for models including midlife occupational complexity, physical activity, and socioeconomic status (SES).

RESULTS

Higher GCA was associated with lower dementia risk (hazard ratio = 0.865, 95% confidence interval = 0.756 to 0.990). After accounting for GCA, no other measure contributed significantly to dementia risk.

DISCUSSION

Putative reserve effects of education or occupational complexity likely reflect largely downstream effects of prior GCA. From a lifespan perspective on reducing dementia risk, the results may suggest favoring interventions aimed at enhancing cognitive development during childhood when there is substantial brain development as opposed to later‐life cognitive training.

Highlights

• Education and GCA serve as indices of cognitive reserve.

• Education is a modifiable risk factor that is associated with dementia risk.

• The effect of higher education on reduced dementia risk is not directly causal.

• Education is largely a downstream effect of prior level of GCA.

• Increasing GCA during childhood may be more effective than later cognitive training.

1. BACKGROUND

Cognitive reserve and related measures are often reported as protective against risk for later‐life cognitive decline or dementia. ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ These include intelligence or general cognitive ability (GCA), education, occupational complexity, engagement in cognitive–intellectual activities, and socioeconomic status (SES). We define cognitive reserve as a person's overall cognitive resources at a given point in time, emphasizing the fact that its level can change across the life course. ¹²

Higher cognitive reserve is assumed to be associated with better brain development. GCA is positively associated with brain volume and cortical surface area. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ Education provides an environment that fosters brain development and neural enhancement, allowing for cognitive enhancement manifested as higher GCA. ¹⁸ , ¹⁹ Enhanced brain development associated with higher GCA, that is, higher cognitive reserve, is thought to be protective against aging‐related atrophy, neural inefficiency, brain disease or damage, and cognitive decline. This presumed protection may foster cognitive resilience, that is, enabling someone to perform well cognitively despite the level of atrophy or disease. ¹²

In most of the aforementioned cognitive reserve studies, the interpretation has been that higher education enhances cognitive function, thereby reducing dementia risk. If education drives GCA, then higher education is likely to be causally related to lower dementia risk. That logic would strongly support cognitive training, an education‐related activity. On the other hand, evidence supports brain differences prior to formal education and GCA as a major factor driving educational attainment. ¹⁹ Instead of being causally related to dementia risk, education may be more of a downstream effect of prior GCA. There may thus be a type of reverse causation, not between education and dementia, but between education and GCA. Because early GCA data are often unavailable, the possibility of this reverse causation is not often directly addressed. In studies with relevant data, the direction of causation appears to be primarily from GCA to education. Two studies examined cognitive, but not dementia, outcomes. In middle‐aged American men, we found that with both earlier GCA and education included in the same model, after accounting for occupational complexity, engagement in cognitive‐intellectual activities, and physical activity, only young adult GCA accounted for significant variance in specific cognitive abilities at ages 56 to 66. ¹⁹ In the Lothian Birth Cohort, age 11 GCA accounted for substantially more variance than did education in specific cognitive abilities at ages 70 to 82. ²⁰

Project Talent, a US sample with GCA data prior to completion of high school, examined participants after nearly 60 years; adolescent GCA had a direct effect on cognitive impairment but also on a dementia questionnaire. ²¹ With GCA scores prior to the completion of education, the authors were also able to show that education, but not occupational complexity, partially mediated those outcomes. Two Mendelian randomization studies have shown that genes for GCA predicted AD after controlling for education‐related genes; genes for education did not predict AD after controlling for GCA‐related genes. ²² , ²³ We are aware of only three studies that examined this issue with respect to dementia diagnostic outcomes using actual GCA and education phenotypes. In these studies, young adult GCA was again a predictor of dementia over and above the effect of education in Danish, Swedish, and Norwegian cohorts. ²⁴ , ²⁵ , ²⁶

RESEARCH IN CONTEXT

1. Systematic review: Higher education and occupational complexity are often viewed as cognitive reserve indices that reduce risk for dementia. Direction of causation regarding their relationship to prior GCA, another cognitive reserve measure, is rarely examined because prior cognitive data are often unavailable.

2. Interpretation: Our findings support the view that higher levels of education or occupational complexity do not cause reduced dementia risk. Rather, these associations are largely downstream effects driven primarily by young adult GCA, the impact of which appears to be long‐lasting.

3. Future directions: Education is a modifiable risk factor. Its effect on dementia risk may be maximized by efforts to enhance GCA during childhood and adolescence, when there is substantial brain development. A lifespan perspective thus raises policy‐relevant questions. Interventions typically focus on later‐life cognitive training, but maximizing resources for enhancing cognitive development during childhood and adolescence might be more effective for reducing dementia risk.

The primary goal of the present study was to examine the impact of education versus GCA on dementia risk and consider implications for public policy. We used 52‐year prospective data beginning in early adulthood to assess the impact of cognitive reserve – defined by GCA or education – on risk for dementia. In contrast to the other studies examining dementia diagnosis outcomes, our longer follow‐up avoided a focus on early‐onset dementia. We also included midlife occupational complexity, physical activity, and SES in our models. The inclusion of such measures is important because, for example, the fact that individuals with higher cognitive reserve (i.e., GCA) are more likely to be in more complex occupations makes reverse causation a strong possibility. ¹⁹

2. METHODS

2.1. Participants

Data were utilized from male members of the Swedish Twin Registry (STR) who participated in the Screening Across the Lifespan (SALT) study. ²⁷ Conducted between 1998 and 2002, the SALT study involved the computer‐assisted assessment of 44,919 STR members (men and women combined) on a variety of health‐related outcomes. Of the 20,879 men who participated in SALT, cognitive data at the time of military conscription were available on 16,619 individuals born between 1936 and 1958. Data from the SALT study were linked to Swedish national health registries in order to obtain data on dementia diagnosis and mortality.

2.2. Young adulthood cognitive ability

Beginning in 1901, all Swedish men were required by law to register for military service, with service requirements varying over the course of the 20th century. Cognitive evaluations were initiated in the 1940s and assessed the domains of inductive reasoning, verbal comprehension, spatial visualization, and technical comprehension. ²⁸ In the present study, data were utilized from three conscription cohorts that were administered variations of the Swedish Enlistment Battery. Cohort 1 was evaluated between 1954 and 1958, Cohort 2 between 1959 and 1968, and Cohort 3 in or after 1969. Age at the time of these cognitive evaluations was 18 years. Scores on individual subtests were converted to a nine‐point Gaussian scale and summed to create a single score. Here we refer to this measure as young adult GCA. Correlations among the subtests within each of the three cohorts varied little (Cronbach alpha range 0.781 to 0.848). Small but significant mean level differences between the cohorts were observed, such that Cohorts 1 and 2 performed worse than Cohort 3. Differences ranged from 0.07 standard deviation (SD) between Cohorts 1 and 3 (p = 0.0045) to 0.11 SD between Cohorts 2 and 3 (p < 0.0001).

2.3. Lifetime education

Level of lifetime education was obtained at the time of the SALT interview and was coded according to a modified version of the International Standard Classification of Education (ISCED) system. ²⁹ For the present analyses, participants with a primary school education or less were given a score of 1. Those with a lower secondary education were given a score of 2, while an upper secondary education was coded as 3. Participants who reported post‐secondary, non‐tertiary education or those who reported short‐cycle tertiary education (e.g., associate's degree) were coded as 4. Finally, participants who reported the equivalent of a bachelor's degree or higher (e.g., master's degree or doctorate) were coded as 5. In addition to this classification system, and in order to provide consistency with other studies, ²⁴ we collapsed the ISCED scale into a three‐level education variable reflecting individuals with low (primary school or less, lower secondary), middle (upper secondary, post‐secondary non‐tertiary), and high levels (short tertiary, bachelor's degree, post‐bachelor's degree).

2.4. Dementia ascertainment

The ascertainment of Alzheimer's disease and related dementia (ADRD) diagnoses for the STR has been described in detail elsewhere. ³⁰ , ³¹ Briefly, registry‐based diagnostic data in the form of International Classification of Disease (ICD) codes (Editions 8, 9, and 10), were obtained for SALT participants from the Swedish National Patient Register (NPR), Outpatient Register (OPR), Prescribed Drug Register (PDR), and the Cause of Death Register (CDR). All diagnostic data, as well as data on participant mortality, were up to date as of December 31, 2016. Age at dementia diagnosis was inferred from date of hospital discharge, date of filled prescription, or date of death. For a small subset of participants in the present study (N = 90), clinically derived dementia diagnoses were available (i.e., diagnoses based on cognitive screening, formal cognitive testing, neurological examination, and review by diagnostic consensus committee). These data were obtained as part of two subsequent STR studies: The Study of Dementia in Swedish Twins (also known as HARMONY) ³² and the Swedish Adoption/Twin Study of Aging (SATSA). ³³ Both studies utilized the Diagnostic and Statistical Manual of Mental Disorders, Third Edition (III) Revised or DSM‐IV criteria for the diagnosis of dementia, and National Institute of Neurological and Communicative Disorders and Stroke ‐ Alzheimer's Disease and Related Disorders Association criteria for the diagnosis of AD. For these participants, age at onset was estimated based on information gathered during in‐person assessments (e.g., informant report, examination of longitudinal cognitive data). ³⁴

2.5. Occupational complexity, SES, and physical activity

Occupational complexity was assessed using the Occupational Cognitive Requirements Score (OCRS), a composite measure of 10 items that assess the cognitive demands of occupations on a scale of 0 to 7. ³⁵ Scores for the present sample ranged from 2.01 to 4.08. Adult SES was assessed with the International Socio‐Economic Index, which combines income and education to capture the overall status of an occupation. ³⁶ Scores for the present sample ranged from 11.74 to 88.96. Finally, physical activity was assessed at the time of the SALT interview using a 7‐point Likert scale that inquired as to the participant's annual exercise pattern. Scores ranged from 0 (almost never exercise) to 6 (maximum amount of exercise).

2.6. Data analysis

Cox proportional hazards models were conducted using proc phreg in SAS, accounting for clustering of twins within families. For all models, age 50 was designated as the start of time, meaning that all participants had to be dementia free prior to age 50. Diagnosis of all‐cause dementia was defined as the model event, and the estimated age of dementia diagnosis was used as the event time. Participants who did not experience the event (i.e., were not diagnosed with all‐cause dementia during follow‐up) were censored at the age of death, emigration, or administratively censored (December 31, 2016). All continuous measures were standardized to a mean of 0 and a standard deviation of 1 prior to their inclusion in the analyses. We tested the following models with dementia diagnosis as the outcome variable: (1) independent effect of young adult GCA; (2) independent effect of education; (3) combined effects of young adult GCA and education; (4) independent and combined effects of young adult GCA with occupational complexity, midlife SES, and physical activity; (5) combined effects of young adult GCA, education, and midlife SES; and (6) effects of these measures after accounting for mortality.

3. RESULTS

Characteristics of the analytic sample are provided in Table 1. Despite similar birth year ranges, the ADRD subsample was on average 5.4 years older than the non‐ADRD subsample (58.0 vs 52.6, p < 0.001). For common measures of cognitive reserve, the ADRD subsample reported lower levels of education relative to the non‐ADRD participants (p < 0.001) and had lower young adult GCA estimates at the time of military induction (p < 0.001). This latter comparison remained statistically significant after controlling for conscription cohort effects, as determined by the tests administered to the participants. Statistically significant differences were also observed for the measures of adult occupation complexity (p = 0.029), and adult SES (p = 0.005), with the ADRD subsample reporting lower scores in each case. The average length of follow‐up from military induction was 52.2 years (SD = 6.3; range: 33 to 63 years).

TABLE 1.

Demographic characteristics.

	Total sample (N = 16,427)	Non‐ADRD (N = 16,040)	ADRD (N = 387)
Age at SALT interview	52.7 (5.58)	52.6 (5.54)	58.0 (4.23)
Birth year (mean, range)	1947 (1936 to 1958)	1947 (1936 to 1958)	1941 (1936 to 1957)
Education
Primary or less	2714 (19.7%)	2,601 (19.3%)	113 (36.0%)
Lower secondary	4241 (30.8%)	4,159 (30.9%)	82 (26.1%)
Upper secondary	3094 (22.5%)	3,028 (22.5%)	66 (21.0%)
Post‐secondary a	1211 (8.8%)	1,191 (8.9%)	20 (6.4%)
Bachelor's degree or higher	2511 (18.2%)	2,478 (18.4%)	33 (10.5%)
Young adult general cognitive ability	18.96 (6.3)	19.00 (6.3)	17.58 (6.5)
Conscription cohort
1954 to 1958	2878 (17.5%)	2,671 (16.7%)	207 (53.5%)
1959 to 1968	7687 (46.8%)	7,535 (47.0%)	152 (39.3%)
1969+	5862 (35.7%)	5,834 (36.4%)	28 (7.2%)
Occupation complexity	3.33 (0.39)	3.33 (0.39)	3.28 (0.40)
Adult SES	46.24 (21.34)	48.37 (21.37)	42.72 (19.78)
Physical activity	3.06 (1.35)	3.06 (1.35)	3.07 (1.36)

Note: N occupation complexity = 12,517; N adult SES = 12,600; N physical activity = 13,746.

Abbreviations: ADRD, Alzheimer's disease and related dementias; SALT, Screening Across the Lifespan; SES, socioeconomic status.

^a The post‐secondary group represents participants with tertiary (associate's degree) and non‐tertiary education (vocational training).

Results of the survival analyses for the independent and combined effects of young adult GCA and education and for the combined effects of GCA, education, and SES are reported in Table 2. The mean age at estimated dementia diagnosis was 70.2 years (SD = 6.3, range: 51 to 81). Independently, each measure was associated with dementia risk, and in the direction that would be expected based on the literature. Higher GCA at military conscription was associated with 15.7% reduced dementia risk. Similarly, when lifetime education was utilized as a continuous measure, higher education was associated with 11.8% reduced dementia risk. When education was used as a categorical measure, the participants with the lowest and middle levels had 56.9% and 69.5% greater dementia risk, respectively, relative to the participants with the highest level.

TABLE 2.

Association of lifetime education attainment, midlife socioeconomic status, and dementia risk with and without the effects of young adult general cognitive ability.

	Estimate (SE)	HR (95% CI)	P
Independent effects a
Young Adult GCA	−0.171 (0.058)	0.843 (0.752; 0.944)	0.0032
Continuous ISCED	−0.126 (0.056)	0.882 (0.790; 0.984)	0.0249
Categorical ISCED
Lowest	0.450 (0.188)	1.569 (1.085; 2.268)	0.0166
Middle	0.528 (0.201)	1.695 (1.143; 2.512)	0.0086
Highest	Ref	1	–

Combined effects (two variables) a	Estimate (SE)	HR (95% CI)	P
Young Adult GCA	−0.145 (0.069)	0.865 (0.756; 0.990)	0.0357
Continuous ISCED	−0.050 (0.065)	0.951 (0.837; 1.081)	0.4436
Young Adult GCA	−0.171 (0.066)	0.843 (0.741; 0.960)	0.0101
Categorical ISCED
Lowest	0.231 (0.208)	1.260 (0.838; 1.895)	0.2665
Middle	0.450 (0.205)	1.568 (1.049; 2.345)	0.0283
Highest	Ref	1	–
Combined effects (three variables) b
Young adult GCA	−0.153 (0.072)	0.859 (0.746; 0.988)	0.0339
Continuous ISCED	−0.008 (0.081)	0.992 (0.846; 1.163)	0.9202
Midlife socioeconomic status	−0.068 (0.078)	0.935 (0.802; 1.089)	0.3860

Note: Analyses are restricted to participants with age of dementia onset or age at last assessment greater than or equal to 50.

^a Total N = 13,771 (Events = 314; Censored = 13,457).

^b Total N = 12,571 (Event N = 286; Censored N = 12,285).

Abbreviations: CI, confidence interval; GCA, general cognitive ability; HR, hazard ratio; ISCED, International Standard Classification of Education; SE, standard error.

In the combined effects models, where the effects of young adult GCA and education were simultaneously tested and thus mutually adjusted for, the effect of young adult GCA on dementia risk was comparable to that in the independent effect model and remained statistically significant (higher GCA associated with 13.5% reduced risk), while the effect of education was notably attenuated (higher education associated with 4.9% reduced risk [non‐significant]). In a combined model where education was used as a categorical variable, dementia risk for those with the lowest level of education was attenuated and no longer statistically significant (26.0% increased risk relative to those with the highest level), whereas those with the middle level continued to show an elevated dementia risk similar to that in the independent effect model (56.8% increased risk). The results were similar in a combined model that simultaneously included midlife SES in addition to young adult GCA and education. Higher GCA was associated with 14.1% reduced risk of dementia. Education and SES were associated with 0.8% and 6.5% reduced risk, respectively (both non‐significant).

Similar analyses were conducted for measures of occupational complexity, SES in adulthood, and physical activity at the time of the SALT assessment. Results for these models are presented in Table 3. In the independent effects models (i.e., not controlling for young adult GCA), only adult SES was found to have a statistically significant association with ADRD risk (higher SES associated with 13.9% reduced risk). Neither occupational complexity nor physical activity showed statistically significant associations with ADRD risk, though estimates were in the expected direction. In the combined effects model, the association of adult SES with ADRD risk was reduced and no longer statistically significant (higher SES associated with 7.0% reduced risk). The effects of occupational complexity and physical activity remained comparable and not statistically significant.

TABLE 3.

Association of lifestyle measures with dementia risk with and without effects of young adult general cognitive ability.

	Estimate (SE)	HR (95% CI)	P
Independent Effects
Occupational complexity	−0.105 (0.058)	0.900 (0.803; 1.09)	0.070
Adult SES	−0.1496 (0.059)	0.861 (0.767; 0.967)	0.011
Physical activity	−0.112 (0.060)	0.894 (0.795; 1.004)	0.059

Combined Effects	Estimate (SE)	HR (95% CI)	P
Young adult GCA	−0.179 (0.064)	0.837 (0.738; 0.948)	0.005
Occupational complexity	−0.043 (0.063)	0.958 (0.846; 1.084)	0.496
Young adult GCA	−0.163 (0.066)	0.850 (0.746; 0.968)	0.014
Adult SES	−0.073 (0.064)	0.930 (0.820; 1.054)	0.255
Young Adult GCA	−0.177 (0.0.59)	0.838 (0.747; 0.940)	0.003
Physical Activity	−0.113 (0.060)	0.894 (0.796; 1.004)	0.059

Note: Sample sizes differed across outcome measures: occupational complexity total N = 12,517 (events = 288; censored = 12,229); adult SES total N = 12,600 (events = 288; censored = 12,312); physical activity total N = 13,746 (events = 314; censored = 13,432). Analyses are restricted to participants with age of dementia onset or age at last assessment greater than or equal to 50.

Abbreviations: CI, confidence interval; GCA, general cognitive ability; HR, hazard ratio; ISCED, International Standard Classification of Education; SE, standard error.

As a check on the extent to which early deaths may have affected the results, we also tested models after removing participants who died prior to 2016 without a dementia diagnosis. In a combined model, higher young adult GCA and continuously measured higher education were significantly associated with 13.8% and 8.8% reduced mortality risk, respectively (Table S1). Despite those associations, however, accounting for mortality did not lead to any meaningful changes in the dementia risk results. In a combined model accounting for mortality, young adult GCA was significantly associated with 14.2% reduced risk of dementia; education was associated with 5.6% reduced risk (non‐significant) (Table S2). Finally, Table S3 shows that results for occupational complexity, midlife SES, and physical activity remained comparable after accounting for mortality.

4. DISCUSSION

4.1. GCA versus education

The link between higher educational attainment and reduced risk for dementia appears to be largely explained by a form of reverse causation. When controlling for young adult GCA, the effect of education on dementia outcomes was substantially attenuated, but after controlling for educational attainment, higher young adult GCA remained robustly associated with lower risk for dementia. Between the two, the causal factor linked to reduced dementia risk is primarily young adult GCA while higher educational attainment appears largely to be a downstream effect of higher GCA. These results were largely unchanged after accounting for midlife SES and mortality.

Our results are consistent with similar prospective studies that examined later‐life cognitive function or dementia, including American and Danish samples as well as another Swedish sample that examined early‐onset dementia. ¹⁹ , ²¹ , ²⁴ , ²⁵ Three of these studies, including the present study, used testing at military induction as the measure of young adult GCA, thus including men only. However, essentially the same result has also been shown in a mixed‐sex sample. ²¹ In addition, in Mendelian randomization analyses of GCA and education in two mixed‐sex samples, GCA, but not education, remained a causal predictor when both were in the models. ²² , ²³ A strength of Mendelian randomization is that genetic associations typically rule out any possibility of reverse causation because an outcome cannot cause the genes. But some complex biopsychosocial phenotypes may, to some extent, be downstream effects of other phenotypes. Indeed, these Mendelian randomization studies are consistent with the idea that some genes that influence GCA in turn influence educational attainment. One study did find that both GCA and education were independently associated with later‐life cognitive function at relatively similar magnitudes. ³⁷ Inclusion of a polygenic score for education in their models, which the authors noted was robustly correlated with adolescent GCA, may, however, raise the question of whether it reduced variance accounted for by GCA.

It is also true that more education increases GCA, largely earlier in life and leveling out by early adulthood. ¹⁹ , ³⁸ Education likely provides cognitive stimulation for the rapidly developing brain earlier in life, but the direction of effect between GCA and education is much more in the other direction once early adulthood is reached. ¹⁹ The Project Talent results seem consistent with this line of reasoning, with a direct effect of adolescent GCA and an indirect effect of education on dementia, that is, education partially mediating the effect of GCA on outcomes. ²¹ Viewing GCA as an index of cognitive reserve, this result would be consistent with the idea that education increases cognitive reserve primarily during periods of substantial brain development and then tends to reach an asymptote. ³⁹ In a 50‐year follow‐up of World War II veterans, young adult GCA accounted for more variance in late‐life cognitive status than education (20.6% vs 16.7%), but a GCA‐by‐education interaction showed that both combined accounted for 4.2% additional variance over GCA alone ⁴⁰ . Other factors (e.g., personality traits ⁴¹ , ⁴² ) may also influence educational attainment and later‐life functioning.

Cognitively enriching activities, including having a more complex occupation, have been associated with reduced risk for dementia, but such factors also appear to be downstream, rather than directly causal, effects of GCA. ¹⁹ , ⁴³ In our study as well as in others, ¹⁹ , ²¹ the effects of such factors were attenuated after accounting for GCA. Enriching activities are not randomly assigned, and individuals with higher GCA may be more likely to seek out intellectually stimulating activities. Similarly, a large twin study showed that genetic influences on GCA increased from childhood to young adulthood, likely due to genetic predisposition to select particular environments. ⁴⁴ Our results are consistent with such factors being much more a function of earlier cognitive reserve than mechanisms for enhancing later cognitive reserve, a perspective with very different implications for intervention.

This study has strengths and limitations. The availability of measured young adult GCA, which is relatively unique in older samples, allowed for rigorous testing of hypotheses relevant to cognitive reserve and reverse causation. The STR participants have also been rigorously followed for data on dementia diagnoses, age of diagnosis, and mortality from multiple Swedish national registries. In contrast to several similar studies of dementia diagnosis outcomes, our 52‐year follow‐up meant that we were not primarily limited to early onset dementia. The present study also accounted for three midlife factors – occupational complexity, SES, and physical activity – that might affect dementia risk. Although the diagnostic data were not derived using current biomarker‐based research standards, ⁴⁵ this should not undercut the data presented here, which have been curated in the decades since the original SALT study. The all‐male, ethnically homogeneous sample (all participants from Sweden) may limit generalizability. However, as noted, similar effects have been observed in mixed‐sex samples.

4.2. Implications for intervention and public policy

How one conceptualizes the relationship between education and GCA can have important implications for research on ADRD, interventions to improve cognitive function, and public policy. If education has a causal role in reducing risk for later‐life cognitive decline and dementia, then it makes sense to focus on efforts to improve or increase education. This view may support the value of cognitive training in later life, which is essentially a form of education as it involves systematic instruction and practice. Efforts at late life cognitive interventions have often had small effects and transfer of training has been inconsistent. ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ Some well‐designed cognitive intervention studies with random assignment to conditions have shown that cognitive training can improve brain and cognitive function or slow cognitive decline. ⁴⁸ , ⁵¹ , ⁵² These interventions tend to require extensive training, a degree of time and effort that many older may not be willing or able to invest. However, if they are effective, they would be valuable even if only a small proportion of older adults would engage in such programs.

On the other hand, our findings may suggest that we might achieve greater overall effect through improvements in prenatal and childhood health that affect early brain health and development (cf. Bratsberg et al. ²⁶ ). Low birthweight and childhood poverty or low SES have substantial effects on brain development and GCA. ⁵³ , ⁵⁴ For example, American children who are well below the poverty line already score lower on cognitive tests, and such scores are predictive of adult functioning and achievement. ⁵³ Moreover, while we commonly look at smaller hippocampal volumes as a risk indicator for Alzheimer's disease, these same children already exhibit smaller hippocampal volumes than other children. ⁵³ Thus, access to and quality of health and education – especially during childhood – are of great importance for maximizing one's potential GCA and, hence, one's level of cognitive reserve. ¹⁸ , ⁵⁵ Such access early in life seems particularly important given that increases in GCA with more years of education appear to reach an asymptote by very early adulthood. ¹⁹ , ⁵⁶ Even the Flynn effect – secular increases in intelligence with increased education – has also shown signs of a leveling off or even partial reversal. ⁵⁷ , ⁵⁸ , ⁵⁹

Although our focus is primarily on aging and dementia, it may be that shifting the balance more toward childhood health and education will ultimately have a greater impact on reducing risk for later‐life cognitive decline and dementia. Gains in childhood and adolescence have lifelong effects. Moreover, the public policy issue is relevant to diversity issues beyond just old versus young. For example, with respect to education quality in the United States, structural racism may play a substantial role. In one large‐scale US study, the proportions of Blacks and Whites in the lowest third for educational quality were essentially mirror images – 76.2% to 86.1% for Blacks versus 20.8% to 23.3% for Whites. ⁵⁵ The build‐up of cognitive reserve that may result from high‐quality education is likely to be meaningfully reduced in a large proportion of minority children. We have also noted that effective late‐life cognitive interventions require considerable time and intensity.

5. CONCLUSION

In sum, the findings of this study indicate that higher young adult GCA is associated with reduced risk for dementia, even after accounting for lifetime educational attainment. Young adult GCA thus constitutes a stronger index of cognitive reserve than education, and it provides some resilience against factors that cause dementia. For the most part, the evidence also indicates that what may appear to be causal effects of education, occupational complexity, or adult SES are, to a large extent, accounted for by reverse causation, that is, they are primarily downstream effects of earlier GCA. Consequently, examining the relationship of these factors to dementia outcomes without accounting for earlier cognitive ability runs the risk of invalid inferences about causality. This conclusion should not be taken to mean that these other factors are unimportant. Rather, our results, along with other findings on the relationship between intelligence and education, suggest the importance of a lifespan perspective. For those of us who study aging and dementia, the focus is typically on later life, but maximizing the benefit that can be gained from early education may be one of the most effective strategies for reducing dementia risk.

AUTHOR CONTRIBUTIONS

The original study design and data collection were conceived and developed by Nancy L. Pedersen and Margaret Gatz. The analysis was conceived and the manuscript written by William S. Kremen and Matthew S. Panizzon. Matthew S. Panizzon analyzed the data. Malin Ericsson and Ida K. Karlsson facilitated acquisition of the conscription data and provided expertise in analysis and interpretation of those data as well as ADRD diagnoses and other measures from the STR data. All authors interpreted the data. All authors contributed to revising/editing the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

All human subjects provided informed consent. The study was approved by local Institutional Review Boards and was performed in accordance with the ethical standards laid down by the 1964 Declaration of Helsinki and its later amendments.

Supporting information

Supporting information

Supporting information

Kremen WS, Ericsson M, Gatz M, et al. A lifespan perspective on cognitive reserve and risk for dementia. Alzheimer's Dement. 2025;21:e70176. 10.1002/alz.70176