- Email: [email protected]
Lifecourse epidemiology matures: Commentary on Zhang et al. “Early-life socioeconomic status, adolescent cognitive ability, and cognition in late midlife”
Social Science & Medicine xxx (xxxx) xxxx
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Lifecourse epidemiology matures: Commentary on Zhang et al. “Early-life socioeconomic status, adolescent cognitive ability, and cognition in late midlife” Amal Harratia, M. Maria Glymourb,∗ a b
Division of Primary Care and Population Health, Department of Medicine, Stanford University, Stanford, 94305, USA Department of Epidemiology and Biostatistics, University of California, San Francisco, 94158, USA
A R T I C LE I N FO
A B S T R A C T
Keywords: Lifecourse Education Dementia Research methods Triangulation
The effect of education on late life cognition has attracted substantial attention in lifecourse epidemiology, in part because of its relevance for understanding the effect of education on dementia. Although numerous studies document an association between education and later life cognition, these studies are potentially confounded by early life socioeconomic position and cognition. Good measures of these early life constructs are rarely available in data sets assessing cognition in late life. A further body of evidence has taken advantage of compulsory schooling law (CSL) instrumental variables (IV), although these estimates have been criticized based on questions about the validity of CSL IVs. In this issue of the Journal, Zhang et al. took advantage of the Wisconsin Longitudinal Study to control for both prospectively measured adolescent IQ and early life socioeconomic status in an analysis evaluating the effect of education on cognitive scores in late middle age (Zhang et al., 2019; IN THIS ISSUE). Their results indicate a moderate effect of each additional year of education on later life cognition, of approximately 0.1–0.15 standard deviations per year of schooling. These estimates are remarkably aligned with findings from prior observational designs and from the CSL IV studies. Although criticisms of any individual study are plausible, this new study complements the body of prior evidence to provide compelling evidence for the benefits of education on late life cognition.
In this issue of the Journal, Zhang and colleagues provide a careful analysis integrating measures of socioeconomic status, health, and cognition across the lifecourse to examine influences on mid-life cognitive function (Zhang et al., 2019, IN THIS ISSUE). Using prospective measures of education and cognitive ability measured at age 17 (the Hammond IQ test) from the Wisconsin Longitudinal Study (WLS), the authors find that both educational attainment and adolescent cognitive ability have direct effects on late-life cognition, with a standardized coefficient of adult cognition roughly twice that of education. They also find that childhood SES has small direct effects on adult cognition but larger indirect effects via adolescent cognitive ability and educational attainment. The term “lifecourse” is often aspirational in lifecourse epidemiology, because few studies have detailed measures at multiple developmental stages. As a result, disentangling the consequences of sequentially correlated characteristics has been challenging. In the study of determinants of adult cognitive function, measures of childhood
∗
cognitive ability are often absent from studies and thus remain an important confounder in understanding lifecourse influences on later-life cognition. From this perspective, the Zhang results provide valuable new information. The WLS fills an important gap by providing highquality, prospective data from adolescence through late midlife, including a measure of adolescent IQ. Zhang's work is a valuable addition to the small number of previous studies reporting on associations between adolescent (McGurn et al., 2008; Kremen et al., 2019; Whalley et al., 2000; Whalley et al., 2005; Foverskov et al., 2019) or adult (Schmand et al., 1997) cognitive ability and later-life cognition. It is also one of the few to draw from data in the United States. Zhang's study also has the advantage that their adolescent cognitive measure is clearly measured at a time that is contemporaneous with or prior to the completion of schooling, a major methodologic advantage. Zhang's results further our understanding about the role of education in later life health and cognition. Their results provide concrete estimates for evaluating confounding by childhood IQ in the study of
DOI of original article: https://doi.org/10.1016/j.socscimed.2019.112575 Corresponding author. Department of Epidemiology and Biostatistics, University of California, San Francisco, 550 16th Street, San Francisco, CA, 94158, USA. E-mail address: [email protected] (M.M. Glymour).
https://doi.org/10.1016/j.socscimed.2019.112645 Received 21 October 2019; Accepted 26 October 2019 0277-9536/ © 2019 Elsevier Ltd. All rights reserved.
Please cite this article as: Amal Harrati and M. Maria Glymour, Social Science & Medicine, https://doi.org/10.1016/j.socscimed.2019.112645
Social Science & Medicine xxx (xxxx) xxxx
A. Harrati and M.M. Glymour
to have an important impact on lifetime cumulative incidence of dementia. Small delays in age of dementia onset lead to large reductions in population burden (Brookmeyer et al., 2007). Taken together, this body of work provides compelling evidence that social factors in childhood and adolescence can be leveraged to help reduce the growing global burden of dementia. Despite the strength of evidence for a causal effect of education in promoting late life cognition, many important questions remain. We do not have a mechanistic understanding of how education is protective, and therefore we have an incomplete understanding of whom it protects. Zhang et al.‘s study suggests both direct effects of education on cognition, as well as a pathway for the health benefits conferred by higher adult socioeconomic position (Turrell et al., 2002; Kaplan and Gavin, 2001; Luo and Waite, 2005; Marden et al. 2017). This is particularly pertinent because a major limitation of the analyses, as acknowledged by the authors, is the relative homogeneity of the WLS sample. How do the relative roles of family SES, education, and cognition differ across racial groups? The authors provide strong evidence for the importance of education in this population of White Americans living in Wisconsin. However, African-Americans born in 1939 were living under Jim Crow laws in the U.S., with a largely segregated educational system and an entirely different set of early childhood challenges (Glymour and Manly, 2018; Manly et al., 2003). As such, it's plausible the role of childhood family and educational experiences differ for Black Americans of that cohort relative to white Americans, and that these differences may shape later-life trajectories of later-life health in distinct ways. Further, Black men in particular may have encountered discrimination that prevented them from reaping the financial and health benefits of gains (Liu et al., 2015; Pollack et al., 2008; Vable et al., 2019). The role of adult SES in dementia risk may be stronger in Black Americans relative to White Americans (Yaffe et al. Barnes et al.; Luo and Waite, 2005), potentially because childhood SES does not so reliably equate to adult SES among Blacks. Moreover, the APOE-e4 genetic allele appears to confer a lower risk to Black Americans relative to Whites (Farrer et al.), despite higher overall disease risk being higher, suggesting environmental and social factors may impact disease risk differently across racial groups. Moreover, Zhang's mediation analysis relies on strong, unverifiable, and likely at least partially incorrect assumptions. These assumptions are shared with nearly all prior work in this area. In particular, mediation analyses are systematically biased if there are unmeasured confounders of either the exposure-outcome association or the mediatoroutcome combination. Measurement error in the mediator leads to underestimates of the indirect effects. Despite a more careful accounting of lifecourse factors, the WLS measures of childhood SES and adult health are still vulnerable to confounding and measurement error. As a result, their results do not firmly establish the causal role of childhood SES; these estimates may be confounded by unmeasured prior factors. Similarly, their results do not establish that the role of adult health as a mechanism is very small, since only a very imprecise measure of adult health is used in the model. The multiple biases make the net bias difficult to anticipate but unmeasured confounding of the mediator-outcome association would typically underestimate the direct effects of an exposure. In conclusion, Zhang's work provides an important improvement in the quality of evidence that increases in education can reduce future risk of late life cognitive impairment and dementia. After many years of incomplete and imperfect studies on the effects of education on late life cognition, a coherent story is beginning to emerge. Complementing prior research with evidence from a longitudinal, prospective study incorporating careful control for potential early life confounders, Zhang's findings provide compelling evidence for a substantial benefit of education on late-life cognitive function. Future directions include better understanding of heterogeneity of effects across racial/ethnic and other sociodemographic characteristics, confirmation that the cognitive reserve benefits translate to meaningful reductions in
later-life cognitive function as well as other research questions. These estimates are likely to be valuable; for example, they could be used to guide sensitivity analyses for future studies that do not have comparable measures. Most importantly, the Zhang results corroborate the importance of education for cognition in this sample of late-middleaged adults. Zhang et al.'s findings complement a robust literature on the long term effects of education on later-life cognitive outcomes. Conventional observational studies have estimated effects of education on general cognition with correlations of approximately 0.30 (Opdebeeck et al., 2016), but these studies rarely incorporate control for childhood IQ as Zhang was able to do. Yet, without careful control for childhood IQ, these observational studies leave unclear whether educational attainment simply proxies for higher childhood cognitive ability. Individuals with better brain development in childhood may excel and advance in school, and it is that better brain health—and not years of schooling—that may provide protective effects against the loss of later-life function. Thus, the relationship between later-life cognitive function and adolescent cognitive ability may reflect reverse causality–that is, high cognitive ability at a young age drives those individuals to seek additional education and life experiences, and better health behaviors (Park, 2019; Kremen et al., 2019). In lieu of direct measures of cognitive ability, researchers have addressed this confounding by adopting quasi-experimental approaches, often based on changes in compulsory schooling laws as instrumental variables for the health effects of education (Hamad et al., 2018; Lleras-Muney, 2005). A meta-analysis of CSL based studies of the effect of education on IQ suggested approximately a 0.14 SD effect per year of education, although that meta-analysis considered cognitive outcomes across the lifecourse (Ritchie and Tucker-Drob, 2018). Assuming a standard deviation of education of 2–3 years in the contributing studies, the Opdebeeck observational meta-analysis equates to a 0.1 to 0.15 SD benefit to general cognition in later life per year of education. Zhang et al.‘s estimate of the total effect of education on late life cognition are nearly identical (correlation of 0.34, or about 0.15 SD per year of education). In other words, all three study types: observational studies without IQ controls, quasi-experimental studies based on CSLs, and now this observational study with control for adolescent IQ, deliver very similar effect estimates. Although there are many important future directions, the core result that increasing education increases late life cognitive functioning – by around 0.1 to 0.15 SD per year of education – seems quite robust. This is extremely good news for the future of late life cognitive impairment and dementia in the US and globally. Dementia affects an estimated 36 million people worldwide, a number projected to increase rapidly due to population aging (Prince et al., 2016). Prior meta-analyses confirm a strong association between education and dementia risk (Meng and D'Arcy, 2012): low versus high education is associated with a 2.6 fold increase in prevalence and a 1.9 fold increase in incidence of dementia. Similar to the literature on cognitive outcomes overall, interpretation of the dementia research has been uncertain because of potential confounding, in particular by childhood IQ. Zhang et al.'s results indicating that confounding by childhood IQ is an unlikely explanation for the overall effect of education on late life cognition helps validate the causal effect of education on dementia because the effect on dementia is thought to be mediated by cognitive reserve or premorbid cognitive level. Studies assessing the effects of education on later-life cognition show strong relationships with cross-sectional levels of cognition but mixed evidence about the role of education in cognitive decline or change (Marden et al., 2017; Alley et al., 2007; Meng and D'Arcy, 2012; Seblova et al., 2019). The brain reserve capacity threshold theory (Satz, 1993; Stern, 2006; 2003) posits that education permits for better performance on neuropsychological cognitive testing but does not prevent the physiological brain insults that lead to decline. Even if education only fosters high levels of brain reserve, without influencing progressive deterioration, it is likely 2
Social Science & Medicine xxx (xxxx) xxxx
A. Harrati and M.M. Glymour
dementia risk and improved quality of life, and better understanding of which characteristics of education matter for subsequent cognitive benefits.
ethnically diverse elders. J. Clin. Exp. Neuropsychol. 25 (5), 680–690. Marden, J.R., et al., Oct. 2017Marden, n.d. Contribution of socioeconomic status at 3 lifecourse periods to late-life memory function and decline: early and late predictors of dementia risk. Am. J. Epidemiol. 186 (7), 805–814. https://doi.org/10.1093/aje/ kwx155. McGurn, B., et al., Sept. 2008McGurn, n.d. Childhood cognitive ability and risk of lateonset alzheimer and vascular dementia. Neurology 71 (14), 1051–1056. https://doi. org/10.1212/01.wnl.0000319692.20283.10. Meng, Xiangfei, D'Arcy, Carl, 2012. Education and dementia in the context of the cognitive reserve hypothesis: a systematic review with meta-analyses and qualitative analyses. PLoS One 7 (6), e38268. Opdebeeck, C., et al., Jan. 2016Opdebeeck, n.d. Cognitive reserve and cognitive function in healthy older people: a meta-analysis. Aging Neuropsychol. Cognit. 23 (1), 40–60. https://doi.org/10.1080/13825585.2015.1041450. Park, D.C., Feb. 2019Park, n.d. Cognitive ability in old age is predetermined by age 20 y. Proc. Natl. Acad. Sci. 116 (6), 1832–1833. https://doi.org/10.1073/pnas. 1821142116. Prince, Martin, et al., Dec. 2016. Recent global trends in the prevalence and incidence of dementia, and survival with dementia. Alzheimer's Res. Ther. 8 (1). https://doi.org/ 10.1186/s13195-016-0188-8. CrossRef. Ritchie, Stuart J., Tucker-Drob, Elliot M., Aug. 2018. How much does education improve intelligence? A meta-analysis. Psychol. Sci. 29 (8), 1358–1369. https://doi.org/10. 1177/0956797618774253. SAGE Journals. Satz, Paul, 1993. Brain reserve capacity on symptom onset after brain injury: a formulation and review of evidence for threshold theory. Neuropsychology 7 (3), 273–295. https://doi.org/10.1037/0894-4105.7.3.273. APA PsycNET. Schmand, B., et al., Nov. 1997Schmand, n.d. The effects of intelligence and education on the development of Dementia. A test of the brain reserve hypothesis. Psychol. Med. 27 (6), 1334–1337. https://doi.org/10.1017/S0033291797005461. Seblova, Dominika, et al., July 2019. Education and Age-Related Decline in Cognitive Performance: Systematic Review and Meta-Analysis of Longitudinal Cohort Studies. https://doi.org/10.31219/osf.io/4mf9z. osf.io. Stern, Yaakov, 2006. Cognitive reserve and alzheimer disease. Alzheimers Dis. Assoc. Disord. 20 (2), 112–117. Stern, Yaakov, 2003. The concept of cognitive reserve: a catalyst for research. J. Clin. Exp. Neuropsychol. 25 (5), 589–593. https://doi.org/10.1076/jcen.25.5.589.14571. Crossref. Turrell, G., et al., Jan. 2002Turrell, n.d. Socioeconomic position across the lifecourse and cognitive function in late middle age. J. Gerontol.: Ser. Bibliogr. 57 (1), S43–S51. https://doi.org/10.1093/geronb/57.1.S43. Vable, Anusha M., et al., 2019. Differential associations between state-level educational quality and cardiovascular health by race: early-life exposures and late-life health. SSM - Popul. Health 8 (Aug), 100418. https://doi.org/10.1016/j.ssmph.2019. 100418. Crossref. Whalley, L.J., et al., 2000. Childhood mental ability and dementia. Neurology 55 (10), 1455–1459. https://doi.org/10.1212/WNL.55.10.1455. Whalley, Lawrence J., et al., Jan. 2005. Childhood IQ, smoking, and cognitive change from age 11 to 64 years. Addict. Behav. 30 (1), 77–88. https://doi.org/10.1016/j. addbeh.2004.04.014. CrossRef. Zhang, Zhenmei, et al., 2019. Early-life Socioeconomic Status, Adolescent Cognitive Ability, and Cognition in Late Midlife: Evidence from the Wisconsin Longitudinal Study. (in press), IN THIS ISSUE.
Acknowledgements We gratefully acknowledge financial support from NIH/NIA RF1AG056164. References Alley, D., et al., Jan. 2007. Education and cognitive decline in older Americans: results from the AHEAD sample. Res. Aging 29 (1), 73–94. https://doi.org/10.1177/ 0164027506294245. CrossRef. Brookmeyer, Ron, et al., July 2007. “Forecasting the global burden of alzheimer's disease. Alzheimer's Dementia 3 (3), 186–191. https://doi.org/10.1016/j.jalz.2007.04.381. Crossref. Foverskov, Else, et al., July 2019. Socioeconomic position across the life course and cognitive ability later in life: the importance of considering early cognitive ability. J. Aging Health 31 (6), 947–966. https://doi.org/10.1177/0898264317742810. SAGE Journals. Glymour, M.M., Manly, J.J., 2018. Compulsory Schooling Laws as quasi-experiments for the health effects of education: Reconsidering mechanisms to understand inconsistent results. Social sci. Med. 214 (1982), 67–69. Hamad, Rita, et al., 2018. How and why studies disagree about the effects of education on health: a systematic review and meta-analysis of studies of compulsory schooling laws. Soc. Sci. Med. 212 (Sept), 168–178. https://doi.org/10.1016/j.socscimed.2018. 07.016. Crossref. Kaplan, G.A., Gavin, T., et al., Apr. 2001Kaplan and Gavin, n.d. Childhood socioeconomic position and cognitive function in adulthood. Int. J. Epidemiol. 30 (2), 256–263. https://doi.org/10.1093/ije/30.2.256. Kaplan, George A., Ranjit, Nalini, et al., 2008. Lifting Gates–Lengthening Lives: Did Civil Rights Policies Improve the Health of African-American Women in the 1960's and 1970's? Russell Sage. Kremen, W.S., Beck, A., Elman, J.A., Gustavson, D.E., Reynolds, C.A., Tu, X.M., FennemaNotestine, C., 2019. Influence of young adult cognitive ability and additional education on later-life cognition. Proceedings of the National Academy of Sciences 116 (6), 2021–2026. Liu, Sze Yan, et al., June 2015. Historical differences in school term length and measured blood pressure: contributions to persistent racial disparities among US-born adults. Zhu, Shankuan (Ed.), PLoS One 10 (6), e0129673. https://doi.org/10.1371/journal. pone.0129673. Crossref. Lleras-Muney, Adriana, Jan. 2005. The relationship between education and adult mortality in the United States. Rev. Econ. Stud. 72 (1), 189–221. https://doi.org/10. 1111/0034-6527.00329. Crossref. Luo, Y., Waite, L.J., Mar. 2005. The impact of childhood and adult SES on physical, mental, and cognitive well-being in later life. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 60 (2), S93–S101. https://doi.org/10.1093/geronb/60.2.S93. CrossRef. Manly, J.J., Touradji, P., Tang, M.X., Stern, Y., 2003. Literacy and memory decline among
3
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Lifecourse epidemiology matures: Commentary on Zhang et al. “Early-life socioeconomic status, adolescent cognitive ability, and cognition in late midlife” Amal Harratia, M. Maria Glymourb,∗ a b
Division of Primary Care and Population Health, Department of Medicine, Stanford University, Stanford, 94305, USA Department of Epidemiology and Biostatistics, University of California, San Francisco, 94158, USA
A R T I C LE I N FO
A B S T R A C T
Keywords: Lifecourse Education Dementia Research methods Triangulation
The effect of education on late life cognition has attracted substantial attention in lifecourse epidemiology, in part because of its relevance for understanding the effect of education on dementia. Although numerous studies document an association between education and later life cognition, these studies are potentially confounded by early life socioeconomic position and cognition. Good measures of these early life constructs are rarely available in data sets assessing cognition in late life. A further body of evidence has taken advantage of compulsory schooling law (CSL) instrumental variables (IV), although these estimates have been criticized based on questions about the validity of CSL IVs. In this issue of the Journal, Zhang et al. took advantage of the Wisconsin Longitudinal Study to control for both prospectively measured adolescent IQ and early life socioeconomic status in an analysis evaluating the effect of education on cognitive scores in late middle age (Zhang et al., 2019; IN THIS ISSUE). Their results indicate a moderate effect of each additional year of education on later life cognition, of approximately 0.1–0.15 standard deviations per year of schooling. These estimates are remarkably aligned with findings from prior observational designs and from the CSL IV studies. Although criticisms of any individual study are plausible, this new study complements the body of prior evidence to provide compelling evidence for the benefits of education on late life cognition.
In this issue of the Journal, Zhang and colleagues provide a careful analysis integrating measures of socioeconomic status, health, and cognition across the lifecourse to examine influences on mid-life cognitive function (Zhang et al., 2019, IN THIS ISSUE). Using prospective measures of education and cognitive ability measured at age 17 (the Hammond IQ test) from the Wisconsin Longitudinal Study (WLS), the authors find that both educational attainment and adolescent cognitive ability have direct effects on late-life cognition, with a standardized coefficient of adult cognition roughly twice that of education. They also find that childhood SES has small direct effects on adult cognition but larger indirect effects via adolescent cognitive ability and educational attainment. The term “lifecourse” is often aspirational in lifecourse epidemiology, because few studies have detailed measures at multiple developmental stages. As a result, disentangling the consequences of sequentially correlated characteristics has been challenging. In the study of determinants of adult cognitive function, measures of childhood
∗
cognitive ability are often absent from studies and thus remain an important confounder in understanding lifecourse influences on later-life cognition. From this perspective, the Zhang results provide valuable new information. The WLS fills an important gap by providing highquality, prospective data from adolescence through late midlife, including a measure of adolescent IQ. Zhang's work is a valuable addition to the small number of previous studies reporting on associations between adolescent (McGurn et al., 2008; Kremen et al., 2019; Whalley et al., 2000; Whalley et al., 2005; Foverskov et al., 2019) or adult (Schmand et al., 1997) cognitive ability and later-life cognition. It is also one of the few to draw from data in the United States. Zhang's study also has the advantage that their adolescent cognitive measure is clearly measured at a time that is contemporaneous with or prior to the completion of schooling, a major methodologic advantage. Zhang's results further our understanding about the role of education in later life health and cognition. Their results provide concrete estimates for evaluating confounding by childhood IQ in the study of
DOI of original article: https://doi.org/10.1016/j.socscimed.2019.112575 Corresponding author. Department of Epidemiology and Biostatistics, University of California, San Francisco, 550 16th Street, San Francisco, CA, 94158, USA. E-mail address: [email protected] (M.M. Glymour).
https://doi.org/10.1016/j.socscimed.2019.112645 Received 21 October 2019; Accepted 26 October 2019 0277-9536/ © 2019 Elsevier Ltd. All rights reserved.
Please cite this article as: Amal Harrati and M. Maria Glymour, Social Science & Medicine, https://doi.org/10.1016/j.socscimed.2019.112645
Social Science & Medicine xxx (xxxx) xxxx
A. Harrati and M.M. Glymour
to have an important impact on lifetime cumulative incidence of dementia. Small delays in age of dementia onset lead to large reductions in population burden (Brookmeyer et al., 2007). Taken together, this body of work provides compelling evidence that social factors in childhood and adolescence can be leveraged to help reduce the growing global burden of dementia. Despite the strength of evidence for a causal effect of education in promoting late life cognition, many important questions remain. We do not have a mechanistic understanding of how education is protective, and therefore we have an incomplete understanding of whom it protects. Zhang et al.‘s study suggests both direct effects of education on cognition, as well as a pathway for the health benefits conferred by higher adult socioeconomic position (Turrell et al., 2002; Kaplan and Gavin, 2001; Luo and Waite, 2005; Marden et al. 2017). This is particularly pertinent because a major limitation of the analyses, as acknowledged by the authors, is the relative homogeneity of the WLS sample. How do the relative roles of family SES, education, and cognition differ across racial groups? The authors provide strong evidence for the importance of education in this population of White Americans living in Wisconsin. However, African-Americans born in 1939 were living under Jim Crow laws in the U.S., with a largely segregated educational system and an entirely different set of early childhood challenges (Glymour and Manly, 2018; Manly et al., 2003). As such, it's plausible the role of childhood family and educational experiences differ for Black Americans of that cohort relative to white Americans, and that these differences may shape later-life trajectories of later-life health in distinct ways. Further, Black men in particular may have encountered discrimination that prevented them from reaping the financial and health benefits of gains (Liu et al., 2015; Pollack et al., 2008; Vable et al., 2019). The role of adult SES in dementia risk may be stronger in Black Americans relative to White Americans (Yaffe et al. Barnes et al.; Luo and Waite, 2005), potentially because childhood SES does not so reliably equate to adult SES among Blacks. Moreover, the APOE-e4 genetic allele appears to confer a lower risk to Black Americans relative to Whites (Farrer et al.), despite higher overall disease risk being higher, suggesting environmental and social factors may impact disease risk differently across racial groups. Moreover, Zhang's mediation analysis relies on strong, unverifiable, and likely at least partially incorrect assumptions. These assumptions are shared with nearly all prior work in this area. In particular, mediation analyses are systematically biased if there are unmeasured confounders of either the exposure-outcome association or the mediatoroutcome combination. Measurement error in the mediator leads to underestimates of the indirect effects. Despite a more careful accounting of lifecourse factors, the WLS measures of childhood SES and adult health are still vulnerable to confounding and measurement error. As a result, their results do not firmly establish the causal role of childhood SES; these estimates may be confounded by unmeasured prior factors. Similarly, their results do not establish that the role of adult health as a mechanism is very small, since only a very imprecise measure of adult health is used in the model. The multiple biases make the net bias difficult to anticipate but unmeasured confounding of the mediator-outcome association would typically underestimate the direct effects of an exposure. In conclusion, Zhang's work provides an important improvement in the quality of evidence that increases in education can reduce future risk of late life cognitive impairment and dementia. After many years of incomplete and imperfect studies on the effects of education on late life cognition, a coherent story is beginning to emerge. Complementing prior research with evidence from a longitudinal, prospective study incorporating careful control for potential early life confounders, Zhang's findings provide compelling evidence for a substantial benefit of education on late-life cognitive function. Future directions include better understanding of heterogeneity of effects across racial/ethnic and other sociodemographic characteristics, confirmation that the cognitive reserve benefits translate to meaningful reductions in
later-life cognitive function as well as other research questions. These estimates are likely to be valuable; for example, they could be used to guide sensitivity analyses for future studies that do not have comparable measures. Most importantly, the Zhang results corroborate the importance of education for cognition in this sample of late-middleaged adults. Zhang et al.'s findings complement a robust literature on the long term effects of education on later-life cognitive outcomes. Conventional observational studies have estimated effects of education on general cognition with correlations of approximately 0.30 (Opdebeeck et al., 2016), but these studies rarely incorporate control for childhood IQ as Zhang was able to do. Yet, without careful control for childhood IQ, these observational studies leave unclear whether educational attainment simply proxies for higher childhood cognitive ability. Individuals with better brain development in childhood may excel and advance in school, and it is that better brain health—and not years of schooling—that may provide protective effects against the loss of later-life function. Thus, the relationship between later-life cognitive function and adolescent cognitive ability may reflect reverse causality–that is, high cognitive ability at a young age drives those individuals to seek additional education and life experiences, and better health behaviors (Park, 2019; Kremen et al., 2019). In lieu of direct measures of cognitive ability, researchers have addressed this confounding by adopting quasi-experimental approaches, often based on changes in compulsory schooling laws as instrumental variables for the health effects of education (Hamad et al., 2018; Lleras-Muney, 2005). A meta-analysis of CSL based studies of the effect of education on IQ suggested approximately a 0.14 SD effect per year of education, although that meta-analysis considered cognitive outcomes across the lifecourse (Ritchie and Tucker-Drob, 2018). Assuming a standard deviation of education of 2–3 years in the contributing studies, the Opdebeeck observational meta-analysis equates to a 0.1 to 0.15 SD benefit to general cognition in later life per year of education. Zhang et al.‘s estimate of the total effect of education on late life cognition are nearly identical (correlation of 0.34, or about 0.15 SD per year of education). In other words, all three study types: observational studies without IQ controls, quasi-experimental studies based on CSLs, and now this observational study with control for adolescent IQ, deliver very similar effect estimates. Although there are many important future directions, the core result that increasing education increases late life cognitive functioning – by around 0.1 to 0.15 SD per year of education – seems quite robust. This is extremely good news for the future of late life cognitive impairment and dementia in the US and globally. Dementia affects an estimated 36 million people worldwide, a number projected to increase rapidly due to population aging (Prince et al., 2016). Prior meta-analyses confirm a strong association between education and dementia risk (Meng and D'Arcy, 2012): low versus high education is associated with a 2.6 fold increase in prevalence and a 1.9 fold increase in incidence of dementia. Similar to the literature on cognitive outcomes overall, interpretation of the dementia research has been uncertain because of potential confounding, in particular by childhood IQ. Zhang et al.'s results indicating that confounding by childhood IQ is an unlikely explanation for the overall effect of education on late life cognition helps validate the causal effect of education on dementia because the effect on dementia is thought to be mediated by cognitive reserve or premorbid cognitive level. Studies assessing the effects of education on later-life cognition show strong relationships with cross-sectional levels of cognition but mixed evidence about the role of education in cognitive decline or change (Marden et al., 2017; Alley et al., 2007; Meng and D'Arcy, 2012; Seblova et al., 2019). The brain reserve capacity threshold theory (Satz, 1993; Stern, 2006; 2003) posits that education permits for better performance on neuropsychological cognitive testing but does not prevent the physiological brain insults that lead to decline. Even if education only fosters high levels of brain reserve, without influencing progressive deterioration, it is likely 2
Social Science & Medicine xxx (xxxx) xxxx
A. Harrati and M.M. Glymour
dementia risk and improved quality of life, and better understanding of which characteristics of education matter for subsequent cognitive benefits.
ethnically diverse elders. J. Clin. Exp. Neuropsychol. 25 (5), 680–690. Marden, J.R., et al., Oct. 2017Marden, n.d. Contribution of socioeconomic status at 3 lifecourse periods to late-life memory function and decline: early and late predictors of dementia risk. Am. J. Epidemiol. 186 (7), 805–814. https://doi.org/10.1093/aje/ kwx155. McGurn, B., et al., Sept. 2008McGurn, n.d. Childhood cognitive ability and risk of lateonset alzheimer and vascular dementia. Neurology 71 (14), 1051–1056. https://doi. org/10.1212/01.wnl.0000319692.20283.10. Meng, Xiangfei, D'Arcy, Carl, 2012. Education and dementia in the context of the cognitive reserve hypothesis: a systematic review with meta-analyses and qualitative analyses. PLoS One 7 (6), e38268. Opdebeeck, C., et al., Jan. 2016Opdebeeck, n.d. Cognitive reserve and cognitive function in healthy older people: a meta-analysis. Aging Neuropsychol. Cognit. 23 (1), 40–60. https://doi.org/10.1080/13825585.2015.1041450. Park, D.C., Feb. 2019Park, n.d. Cognitive ability in old age is predetermined by age 20 y. Proc. Natl. Acad. Sci. 116 (6), 1832–1833. https://doi.org/10.1073/pnas. 1821142116. Prince, Martin, et al., Dec. 2016. Recent global trends in the prevalence and incidence of dementia, and survival with dementia. Alzheimer's Res. Ther. 8 (1). https://doi.org/ 10.1186/s13195-016-0188-8. CrossRef. Ritchie, Stuart J., Tucker-Drob, Elliot M., Aug. 2018. How much does education improve intelligence? A meta-analysis. Psychol. Sci. 29 (8), 1358–1369. https://doi.org/10. 1177/0956797618774253. SAGE Journals. Satz, Paul, 1993. Brain reserve capacity on symptom onset after brain injury: a formulation and review of evidence for threshold theory. Neuropsychology 7 (3), 273–295. https://doi.org/10.1037/0894-4105.7.3.273. APA PsycNET. Schmand, B., et al., Nov. 1997Schmand, n.d. The effects of intelligence and education on the development of Dementia. A test of the brain reserve hypothesis. Psychol. Med. 27 (6), 1334–1337. https://doi.org/10.1017/S0033291797005461. Seblova, Dominika, et al., July 2019. Education and Age-Related Decline in Cognitive Performance: Systematic Review and Meta-Analysis of Longitudinal Cohort Studies. https://doi.org/10.31219/osf.io/4mf9z. osf.io. Stern, Yaakov, 2006. Cognitive reserve and alzheimer disease. Alzheimers Dis. Assoc. Disord. 20 (2), 112–117. Stern, Yaakov, 2003. The concept of cognitive reserve: a catalyst for research. J. Clin. Exp. Neuropsychol. 25 (5), 589–593. https://doi.org/10.1076/jcen.25.5.589.14571. Crossref. Turrell, G., et al., Jan. 2002Turrell, n.d. Socioeconomic position across the lifecourse and cognitive function in late middle age. J. Gerontol.: Ser. Bibliogr. 57 (1), S43–S51. https://doi.org/10.1093/geronb/57.1.S43. Vable, Anusha M., et al., 2019. Differential associations between state-level educational quality and cardiovascular health by race: early-life exposures and late-life health. SSM - Popul. Health 8 (Aug), 100418. https://doi.org/10.1016/j.ssmph.2019. 100418. Crossref. Whalley, L.J., et al., 2000. Childhood mental ability and dementia. Neurology 55 (10), 1455–1459. https://doi.org/10.1212/WNL.55.10.1455. Whalley, Lawrence J., et al., Jan. 2005. Childhood IQ, smoking, and cognitive change from age 11 to 64 years. Addict. Behav. 30 (1), 77–88. https://doi.org/10.1016/j. addbeh.2004.04.014. CrossRef. Zhang, Zhenmei, et al., 2019. Early-life Socioeconomic Status, Adolescent Cognitive Ability, and Cognition in Late Midlife: Evidence from the Wisconsin Longitudinal Study. (in press), IN THIS ISSUE.
Acknowledgements We gratefully acknowledge financial support from NIH/NIA RF1AG056164. References Alley, D., et al., Jan. 2007. Education and cognitive decline in older Americans: results from the AHEAD sample. Res. Aging 29 (1), 73–94. https://doi.org/10.1177/ 0164027506294245. CrossRef. Brookmeyer, Ron, et al., July 2007. “Forecasting the global burden of alzheimer's disease. Alzheimer's Dementia 3 (3), 186–191. https://doi.org/10.1016/j.jalz.2007.04.381. Crossref. Foverskov, Else, et al., July 2019. Socioeconomic position across the life course and cognitive ability later in life: the importance of considering early cognitive ability. J. Aging Health 31 (6), 947–966. https://doi.org/10.1177/0898264317742810. SAGE Journals. Glymour, M.M., Manly, J.J., 2018. Compulsory Schooling Laws as quasi-experiments for the health effects of education: Reconsidering mechanisms to understand inconsistent results. Social sci. Med. 214 (1982), 67–69. Hamad, Rita, et al., 2018. How and why studies disagree about the effects of education on health: a systematic review and meta-analysis of studies of compulsory schooling laws. Soc. Sci. Med. 212 (Sept), 168–178. https://doi.org/10.1016/j.socscimed.2018. 07.016. Crossref. Kaplan, G.A., Gavin, T., et al., Apr. 2001Kaplan and Gavin, n.d. Childhood socioeconomic position and cognitive function in adulthood. Int. J. Epidemiol. 30 (2), 256–263. https://doi.org/10.1093/ije/30.2.256. Kaplan, George A., Ranjit, Nalini, et al., 2008. Lifting Gates–Lengthening Lives: Did Civil Rights Policies Improve the Health of African-American Women in the 1960's and 1970's? Russell Sage. Kremen, W.S., Beck, A., Elman, J.A., Gustavson, D.E., Reynolds, C.A., Tu, X.M., FennemaNotestine, C., 2019. Influence of young adult cognitive ability and additional education on later-life cognition. Proceedings of the National Academy of Sciences 116 (6), 2021–2026. Liu, Sze Yan, et al., June 2015. Historical differences in school term length and measured blood pressure: contributions to persistent racial disparities among US-born adults. Zhu, Shankuan (Ed.), PLoS One 10 (6), e0129673. https://doi.org/10.1371/journal. pone.0129673. Crossref. Lleras-Muney, Adriana, Jan. 2005. The relationship between education and adult mortality in the United States. Rev. Econ. Stud. 72 (1), 189–221. https://doi.org/10. 1111/0034-6527.00329. Crossref. Luo, Y., Waite, L.J., Mar. 2005. The impact of childhood and adult SES on physical, mental, and cognitive well-being in later life. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 60 (2), S93–S101. https://doi.org/10.1093/geronb/60.2.S93. CrossRef. Manly, J.J., Touradji, P., Tang, M.X., Stern, Y., 2003. Literacy and memory decline among
3