Association of Changes in C-Reactive Protein Level Trajectories Through Early Adulthood With Cognitive Function at Midlife

Abstract

Background and Objectives

Late-life inflammation has been linked to dementia risk and preclinical cognitive decline, but less is known about early adult inflammation and whether this could influence cognition in midlife. We aimed to identify inflammation levels through early adulthood and determine association of these trajectories with midlife cognition.

Methods

We used data from the Coronary Artery Risk Development in Young Adults study to identify inflammation trajectories (C-reactive protein [CRP] level <10 mg/L) over 18 years through early adulthood (age range 24–58) in latent class analysis and to assess associations with cognition 5 years after the last CRP measurement (age range 47–63). Six cognitive domains were evaluated from tests of verbal memory, processing speed, executive function, verbal and category fluency, and global cognition; poor cognitive performance was defined as a decline of ≥1 SD less than the mean on each domain. The primary outcome was poor cognitive performance. Logistic regression was used to adjust for demographics, smoking, alcohol use, physical activity, and APOE 4 status.

Results

Among 2,364 participants, the mean (SD) age was 50.2 (3.5) years; 55% were female, and 57% were White. Three CRP trajectories emerged over 18 years: lower stable (45%), moderate/increasing (16%), and consistently higher (39%). Compared with lower stable CRP, both consistently higher (adjusted odds ratio [aOR] 1.67, 95% CI 1.23–2.26) and moderately/increasing (aOR 2.04, 95% CI 1.40–2.96) CRP had higher odds of poor processing speed; consistently higher CRP additionally had higher odds of poor executive function (aOR 1.36, 95% CI 1.00–1.88). For memory (moderately/increasing aOR 1.36, 95% CI 1.00–1.88; consistently higher aOR 1.18, 95% CI 0.90–1.54), letter fluency (moderately/increasing aOR 1.00, 95% CI 0.69–1.43; consistently higher aOR 1.05, 95% CI 0.80–1.39), category fluency (moderately/increasing aOR 1.16, 95% CI 0.82–1.63; consistently higher aOR 1.11, 95% CI 0.85–1.45), or global cognition (moderately/increasing aOR 1.16, 95% CI 0.82–1.63; consistently higher aOR 1.11, 95% CI 0.85–1.45), no association was observed.

Discussion

Consistently higher or moderate/increasing inflammation starting in early adulthood may lead to worse midlife executive function and processing speed. Study limitations include selection bias due to loss to follow-up and reliance on CRP as the only inflammatory marker. Inflammation is important for cognitive aging and may begin much earlier than previously known.

Introduction

With a growing body of evidence implicating inflammatory mechanisms in the pathogenesis of Alzheimer disease (AD) and other dementias,^1-3 interest in elucidating the role of inflammation as a risk factor of these age-related conditions has increased greatly. Several studies have primarily focused on late-life inflammation,^4-9 with much evidence suggesting an increased risk of dementia and cognitive decline with elevated levels of inflammation in late life.^7-9 If inflammation exposure has long-term effects, exposure earlier in life could have important implications for cognitive aging. However, data are limited regarding earlier life inflammation exposure and whether this could influence cognition in midlife.

Longitudinal studies suggest that brain changes leading to AD and other dementias often take decades to develop and may begin to emerge as early as midlife, but the influence of these earlier changes on cognitive aging is unclear.^10,11 Over the life course, inflammation levels tend to have substantial intraindividual variability, and this variation in levels of inflammation over time may be a critical predictor for cognitive aging.¹² Although studies have reinforced this finding,⁵ many studies on the relationship between inflammation and development of dementia and preclinical cognitive decline have measured inflammation at only one time point,^8,9,13-15 which cannot adequately characterize the cumulative burden of inflammatory activity or how levels of inflammation change over time.

As part of the ongoing Coronary Artery Risk Development in Young Adults (CARDIA) study, we examined the association between early to midlife inflammation trajectories and midlife cognition. We hypothesized that individuals with higher inflammation exposure over these earlier life periods would be at greater risk of poor midlife cognition.

Methods

Study Design and Sample

The CARDIA study is a prospective cohort study of 5,115 White and Black men and women, initially healthy and aged between 18 and 30 years at enrollment in 1985–6. Participants were recruited from 4 US cities (Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA) and completed a baseline visit (year 0) and 8 follow-up visits (years 2, 5, 7, 10, 15, 20, 25, and 30), and the sampling methodology was designed to achieve a balance of each of the 4 sites by race (self-identified as Black, not Hispanic or White, not Hispanic), sex, educational level (high-school degree or less or more than high school), and age (18–24 years or 25–30 years).

Standard Protocol Approvals, Registrations, and Patient Consents

At each visit, participants provided written informed consent, and study protocols were reviewed by institutional review boards from each study site (Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA), the CARDIA Coordinating Center, at the University of Alabama, Birmingham, and the University of California, San Francisco; additional details of the study are provided elsewhere.^16,17

C-Reactive Protein

The systemic inflammatory marker C-reactive protein (CRP) was measured in years 7, 15, 20, and 25, over 18 years. Participants were asked to fast for at least eight hours and avoid smoking and heavy physical activity for at least two hours before each blood sample collection. Plasma for CRP assays was separated from whole blood and shipped on dry ice to a central laboratory for storage. Plasma samples for CRP assays were analyzed using a particle-enhanced immunonephelometric assay with a Siemens Dade Behring BN II Nephelometer in years 7, 15, and 20. The intra-assay coefficients of variation (CoVs) were 2.3%–4.4%, and interassay CoVs were 2.1%–5.7%. In year 25, plasma CRP assays were assessed using a Roche latex-particle enhanced immunoturbidimetric assay kit and Roche Modular P Chemistry analyzer. The intra-assay CoV was 3.7%. Across the two methods, the range was identical (CRP 0.175–1,000 mg/L). These CRP data were used in the analysis to generate inflammation trajectory groups, as specified in the Data Analysis section. We considered using CRP to index inflammation because the use of CRP data for identifying inflammation trajectories has been shown to distinguish groups over time in CARDIA¹⁸ and other cohort studies.⁵

Cognitive Assessment

In year 30 (5 years after the last CRP measurement), trained interviewers administered a battery of 6 cognitive tests: The Rey Auditory Verbal Learning Test assesses verbal memory (we refer to this domain as verbal memory)¹⁹; the Digit Symbol Substitution Test measures processing speed, executive function, and working memory (we refer to this domain as processing speed domain)²⁰; the Montreal Cognitive Assessment evaluates global cognition (we refer to this domain as global cognition domain)²¹; the letter and category fluency tests assess verbal production, semantic memory, phonemic fluency, and language each over one minute (we refer to these domains as verbal and category fluency, respectively)²²; and the Stroop Interference Test evaluates executive function (we refer to this domain as executive function).^23,24 We used the inverse of the Stroop Interference Test to allow for interpretation of better function with higher scores on all cognitive tests.²⁵ We also computed a composite cognitive score. We also computed z-scores for each test and defined poor cognitive performance by a cutoff of ≥1 SD less than the cohort mean for each test, as previously used in CARDIA²⁶ and other population-based studies.²⁷

Covariates

We used covariates as measured in year 25, which was at the last CRP measurement, and 5 years before cognitive assessments were administered (year 30). The covariates considered were participant characteristics, prespecified using relevant literature and our previous work in the CARDIA study.¹⁸ We obtained demographic information from participants based on self-reports. Depression was defined as a score ≥16 on the 20-item Center for Epidemiologic Studies Depression Scale.²⁸ Alcohol use (standardized for different types of alcohol as drinks per week) and cigarette smoking status (current smoker vs current nonsmoker) were assessed by self-report. Physical activity was measured with the CARDIA Physical Activity questionnaire, which queries time spent in 13 physical activities in the past year²⁹ and produces a total score in exercise units, which we used to compare participants with scores ≥/<300 exercise units, approximating the cutoff for recommended levels of physical activity.³⁰ APOE ε4 (presence or not), the high-risk variant of the APOE gene—the gene most commonly associated with increased risk of late-onset AD—was determined from year 7 blood samples using standard techniques.^31,32 Height and weight were measured, and body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared.

Data Analysis

Inflammation trajectories were computed with CRP measured over 18 years (average CRP <10 mg/L in years 7, 15, 20, and 25) using a methodology previously used in CARDIA^18,33 and other population-based studies.³⁴ We used latent class growth analysis, a person-centered approach to modeling that can be used to distinguish groups of individuals based on their probability of following a similar trajectory over time.^35,36 Repeated measurements of CRP were log transformed to account for skewed distributions and then modeled as censored normal using SAS PROC traj³⁶ (eTable 1, for distribution of CRP values). We performed model selection according to recommended procedures such that each trajectory group had to include at least 5% of participants. We fit 2-, 3-, and 4-class models, and in each model for which a given number of trajectories were designated, cubic, quadratic, and linear terms were evaluated to identify trajectory shapes that best fit the data.³⁷ We identified the number of trajectories to retain by evaluating the log Bayes factor, which compares Bayesian information criterion (BIC) values between models and indicates that the more complex model with lower BIC values provides the best model fit. While a 4-trajectory model was associated with further improvement in fit, a one-trajectory model included <5% of participants; therefore, a 3-trajectory model was retained for analyses (eTable 2, for model fit parameters). We performed the trajectory analysis using SAS version 9.4.

Participants with different inflammation trajectories were compared using descriptive statistics. We ran unadjusted and adjusted logistic regression models to examine the association between early-to-midlife inflammation trajectories and midlife cognitive performance. Adjusted models included age, race, sex, education, smoking, alcohol use, physical activity, and APOE 4. The level of significance was set at p < 0.05. Comparisons among the participants with different inflammation trajectories using descriptive statistics and the logistic regression models were conducted in R version 3.4.3.

Data Availability

Anonymized data are available from the CARDIA Coordinating Center (cardia.dopm.uab.edu/contact-cardia). A complete explanation of the National Heart, Lung, and Blood Institute policies governing the data and describing access to the data can be found online (cardia.dopm.uab.edu/study-information/nhlbi-data-repository-data).

Results

Of the 3,039 participants with repeated CRP measurements over 18 years (3 or more visits in years 7, 15, 20, or 25), 134 were excluded because of values of CRP ≥10 mg/L (on >2 consecutive visits because of (1) the possibility that they were from repeat acute inflammatory episodes due to acute illness/injury and (2) the inability to distinguish between elevated inflammation and acute illness/injury or chronic medical conditions with flares).³⁸ This resulted in 2,905 participants for the analysis to identify CRP trajectory groups (eFigure 1, for cohort flow diagram). Participants not included in the analysis to identify CRP trajectory groups had more health risk factors (smoking, lower physical activity score, and depression) and greater prevalence of medical conditions (obesity, hypertension, diabetes, stroke/transient ischemic attack [TIA], and kidney problems) (eTable 3). Overall, our final analytic cohort included 2,364 participants consisting of those from the CRP trajectory group sample who completed at least 1 cognitive test in year 30. Participants not included in the analytic cohort had lower education and were more likely to be Black, male, and smokers and have depression (p < 0.05).

The 3 identified inflammation trajectories reflected overall patterns of lower, moderate, or higher CRP levels and either remained mostly stable or increased from early adulthood to midlife, over 18 years. We labeled the inflammation trajectories as consistently higher, moderate/increasing, and lower stable CRP. Figure 1 shows the inflammation trajectories over 18 years (for individual and group means for the inflammation trajectory groups eFigure 2). Of the 2,364 participants, 39.0% had consistently higher CRP (n = 911), 16.0% had moderate/increasing CRP (n = 381), and 45.0% had lower stable CRP (n = 1,072). The range of CRP values for the inflammation trajectories over 18 years is presented in eTable 4.

Figure 1. Inflammation Trajectories Over 18 Years.

The trajectories of inflammation reflected patterns of lower, moderate, or higher CRP levels and remained stable or increased from early adulthood to midlife, over 18 years. The 3 inflammation trajectories of the 2,364 CARDIA participants were characterized as follows by the year 25 visit: consistently higher (n = 911 [39.0%]) shown in green, moderate/increasing (n = 381 [16.0%]) shown in blue, and lower stable (n = 1,072 [45.0%]) shown in red. CARDIA = Coronary Artery Risk Development in Young Adults; CRP = C-reactive protein.

The Table shows the midlife characteristics of participants by inflammation trajectory group. Most demographics and risk factors of cognitive aging varied by inflammation trajectory group, such that those in the consistently higher CRP group had lower education and were more likely to be Black or female than in the other groups. Smoking and low physical activity were higher in those in the consistently higher CRP group. The proportion of participants with the APOE 4 phenotype was higher in moderate/increasing and lower stable CRP groups. The participant trajectory group did not differ in age or depressive symptoms.

Table.

Characteristics of the 2,364 CARDIA Participants by Inflammation Trajectory Group

Variable	Trajectory			p value
	Lower stable (n = 1,072)	Moderate/increasing (n = 381)	Consistently higher (n = 911)
Age, y	51 (48–53)	50 (47–53)	51 (47–53)	0.81
Female	508 (48.6)	194 (53.0)	571 (64.7)	<0.001
White	721 (67.3)	202 (53.0)	420 (46.1)	<0.001
Education, y	15.6 (2.6)	15.3 (2.7)	14.8 (2.5)	<0.001
Current smoking	107 (10.2)	55 (15.0)	165 (18.7)	<0.001
Alcohol, drinks per week	5 (1–10)	3 (0–9)	3 (0–9)	<0.001
CES-D-Depressiona	155 (14.8)	55 (15.0)	157 (17.8)	0.18
Physical activityb	590 (56.4)	169 (46.2)	317 (35.9)	<0.001
APOE 4c	329 (31.5)	123 (33.0)	232 (26.2)	0.01
BMI, kg/m2	26.8 (4.5)	30.4 (5.6)	33.0 (7.3)	<0.001

Abbreviations: BMI = body mass index; CARDIA = Coronary Artery Risk Development in Young Adults; CES-D = Center for Epidemiologic Studies Depression scale; IQR = interquartile range.

Data are presented as mean (SD), n (%), or median (IQR).

^aSelf-reported depression was measured using the CESD-Depression scale. A score ≥16 was used as the cutoff value as an indication of clinically significant depressive symptoms.

^bPhysical activity measured with the CARDIA Physical Activity History Questionnaire. A score of 300 or higher indicates greater levels of physical activity.

^cAPOE 4 was derived from year 7 blood samples.

Figure 2 displays the odds associated with poor cognitive performance by inflammation trajectory group (for odds ratios and 95% CIs; eTable 5). Compared with lower stable CRP in unadjusted models, having consistently higher or moderate/increasing CRP levels over time was associated with higher odds of poor cognitive performance on almost all domains (for predicted probabilities and 95% CIs derived from unadjusted models; eTable 6). The highest estimated effects were observed for processing speed; the odds of poor performance were >2 times higher on the processing speed domain for both consistently higher and moderate/increasing CRP groups than for the lower stable CRP group. The estimated effects of poor cognitive performance were similar for the executive function domain; the odds were nearly 2 times higher. For all other domains, the odds associated with poor cognitive performance were between 1.4 and 1.5. In fully adjusted analyses (including further adjustment for age, race, sex, education, smoking, alcohol use, physical activity, and APOE 4), the association was attenuated but remained had higher odds of poor performance on processing speed and executive function. However, there was no longer an association observed for verbal memory, fluency, or global cognition.

Figure 2. Unadjusted and Adjusted Odds Associated With Cognitive Performance and Inflammation Trajectory Group.

All year 30 cognitive test results were standardized (z-scores); poor cognitive performance was defined by a cutoff of ≥1 SD lower than the mean for each test. The inverse of the Stroop test was used to allow interpretation of poor cognitive performance on all tests. Unadjusted models are shown in panel A and adjusted models in panel B. Adjusted models included age, race, sex, education, smoking, alcohol use, physical activity, and APOE 4 status. CARDIA = Coronary Artery Risk Development in Young Adults; DSST = Digit Symbol Substitution Test; MoCA = Montreal Cognitive Assessment; RAVLT = Rey Auditory Verbal Learning Test; circles = odds ratios; and error bars = 95% CI. eTable 5 shows results presented as odds ratios with associated 95% CIs.

In a sensitivity analysis, we further adjusted for BMI, which did not significantly alter the results (eTable 5). In addition, there were no significant interactions with race or APOE ε4 for the association between poor cognitive performance by inflammation trajectory group (p > 0.05 for all). Given that inflammation is often assessed at a single time point, we also repeated our main analyses by defining inflammation groups based on the last CRP measurement (year 25) rather than the estimated trajectory groups and found that the strength of the effects was slightly attenuated but similar in direction and pattern (eTable 7).

Discussion

In this prospective cohort study, we found that patterns of consistently higher and moderate/increasing inflammation levels (vs lower stable inflammation) starting in early adulthood were associated with higher odds of poor cognitive function in midlife. These trajectories were associated with slower processing speed and worse executive function after controlling for demographics, lifestyle risk factors, and APOE 4. Participants with a pattern of consistently higher inflammation were most likely to have higher odds of poor cognitive function. There was no association of inflammation trajectory and impairment in memory, fluency, or global cognition.

Our findings are supported by previous prospective studies of inflammation in midlife and late life. These earlier studies indicated that elevated inflammation levels are associated with poor cognitive performance and risk of cognitive decline and dementia in older adults. Results from a large cohort of Japanese American men found that higher CRP levels during middle age (50–58 years) more than doubled the risk of developing dementia by late life (75–85 years).¹⁴ Another study of mostly White elders found that elevated interleukin-6 levels in late adulthood (70–79 years) contributed to poor cognitive performance and late-life cognitive decline.³⁹ Findings were similar in a prospective study of Black and White community-dwelling older adults (mid 70s), in whom higher interleukin-6 and CRP levels were associated with cognitive decline over follow-up.⁸ We focused on early adulthood to midlife and found that inflammation exposure in this earlier stage of the life course is also associated with higher odds of poor cognitive function in midlife. Our results show that a trajectory of consistently higher inflammation, even starting in early adulthood (24–30 years), may contribute to higher odds of poor cognitive function (42–59 years). Moderate/increasing inflammation trajectories may also be associated with higher odds of poor cognition. These findings suggest that inflammation is important for cognitive aging and may begin much earlier in life than previously known.

Our findings of associations with executive function and processing speed are supported by the results from the Whitehall II cohort study, which estimated the effects of midlife inflammation and cognitive function and decline during late middle age. The most robust results indicated that higher midlife CRP levels mostly are associated with higher odds of poor executive function but similarly not worse memory.¹³ Other investigations in slightly older cohorts, such as the ARIC cohort study⁴⁰ and the English Longitudinal Study of Aging,⁷ have also examined inflammation exposure in midlife and risk of cognitive decline later in life but found that memory in late life may be more susceptible to the effects of midlife inflammation relative to other cognitive domains. It is possible that differences in the methods may explain some of this variation in findings, such as the tests used to assess cognitive performance.^4,13 Because inflammation is a risk factor of several cerebrovascular diseases, such as hypertension, high cholesterol, obesity, and diabetes, it is also possible that inflammation may influence cognition through vascular pathways.⁴¹ Results from studies of midlife vascular risk factors, such as hypertension⁴² and high cholesterol,⁴³ suggest that these conditions may be more strongly associated with midlife declines in executive function and processing speed. Indeed, our results observed with processing speed and executive function for both higher and increasing inflammation levels may reflect associations with vascular mechanisms. In a previous CARDIA analysis,¹⁸ we found that elevated inflammation levels through early adulthood were associated with worse midlife brain health, as indexed by lower white matter volume and white matter integrity, brain changes that are most affected by executive function and processing speed.⁴⁰ These findings align with our results and suggest that effects of early to midlife inflammation levels may differentially increase risk of poor midlife cognition by domain, which needs to be clarified in future investigations.

Chronic inflammation may lead to neurodegeneration and death of neurons and impair neurogenesis, the growth of new neurons.^44-46 In animal models, chronic inflammation levels were associated with greater age-related changes in hippocampal structure and function, which may contribute to neurogenerative disease or glial injury.⁴⁶ Other experimental studies suggest that activation of the peripheral inflammatory response may generate neuroinflammatory response (through multiple routes, vagal nerve signaling, and circumventricular organs), which can, in turn, disrupt neural and glial functioning.⁴⁴ Other studies suggest that CRP-mediated impacts on the blood-brain barrier (by increased blood-brain barrier permeability) may be a key mechanism in the association between inflammation and cognitive aging.^47,48 Studies in humans also support a role for inflammation and AD and other neurodegenerative diseases and indicate that inflammation may directly influence cognition through the process of neurodegeneration.³ Imaging studies suggest that elevated inflammation may prospectively predict late-life white matter volume and integrity.⁴⁹ Although some neuropathologic studies indicate that inflammation exposure may interact with amyloid pathways and increase amyloid deposition and plaques in the brain,³ we did not find an interaction between inflammation trajectories and the APOE 4 phenotype. It is possible that inflammation affects cognition through non-APOE 4 pathways or there may be differences in these associations by age.⁴⁰

This study reported an association between inflammation trajectories including three measurements of CRP starting in early adulthood and higher odds of poor cognition in midlife. In this study, we also assessed the strength of the association between inflammation and cognitive function by including only one measurement of CRP (year 25) and observed that the odds of poor midlife cognitive function were attenuated, but in similar direction. These observations align with increasing evidence that indicates that there may be high intraindividual variability in CRP levels over time^5,12 and suggests that multiple assessment periods may be critical for capturing ongoing exposure. Although multiple measurements of CRP were used in this study, the potential of CRP for high intraindividual variability suggests that, ideally, multiple CRP measurements may need to be taken per assessment period (in the context of this study, per “year group”) to derive a more representative annual average. Absent using multiple measurements of CRP to derive an average reading per year category, baseline inflammation levels may not be reliably captured and could potentially explain the mixed findings with cognitive outcomes.

The CARDIA study is a well-characterized, large, and diverse longitudinal cohort, in which we were able to identify early-to-midlife inflammation trajectories and link these trajectories to cognitive function at midlife. We also used a population-based sampling method and accounted for several potential confounders. However, there are limitations to consider. Participant retention was high over the study period, but it is possible that there was some selection bias owing to loss to follow-up. However, we expect this bias would contribute to bias toward the null. We relied on CRP as the marker of inflammation because other inflammatory markers and more direct measures of neuroinflammation are currently unavailable. Apart from the covariates described, we did not assess the contribution of other conditions that are associated with elevated levels of inflammation to our findings (e.g., stroke/TIA hypertension, diabetes, liver disease, multiple sclerosis, kidney disease, and HIV). Participants with elevated levels of inflammation (CRP ≥10 mg/L) were not included in our study, but we could not determine whether their elevated values were from recent illness/injury (possibly indicating a pattern of acute inflammation) or chronic medical condition with flares (possibility indicating a pattern of chronic inflammation).³⁸ It should also be noted that cognitive domains, including those analyzed in this study, are intercorrelated (eFigure 3), and our results should be interpreted given this caveat. Furthermore, some questions may arise regarding the applicability of our main findings (given overlapping 95% CIs in Figure 2), and it is important to note that the extent of significant estimated effects observed (such as processing speed and executive function) may not necessarily suggest a differential contribution to inflammation as compared with the other cognitive domains and our results should be interpreted accordingly. Moreover, because we used covariates as assessed in year 25, some of the covariates analyzed in this study may have occurred after the inflammation exposure, and our results should be interpreted given this caveat. Finally, while our measure of poor cognitive performance, defined as a decrease of greater than 1 SD lower than the mean, is standard in epidemiologic studies of older adults, the clinical significance of poor cognitive function is not well defined during the midlife period and the relevance this early in the life course is unclear.⁵⁰ Thus, further population-based cognition studies will be needed to examine the generalizability of our findings.

Our results indicate that higher or increasing inflammation trajectories may be associated with (or influence) higher odds of poor cognitive function at midlife. Although current public health and prevention efforts mainly focus on late life, our study provides evidence for the need to also target cognitive brain health in middle age. Young and middle-aged adults, especially those who have a trajectory of consistently higher or moderate/increasing inflammation, may represent important subgroups for early monitoring. Additional research is needed to improve early detection of those at the highest risk of poor cognitive performance and to determine effective strategies to delay the process of cognitive aging even earlier in the life course by addressing the drivers of inflammation.

Glossary

AD

Alzheimer disease

BIC

Bayesian information criterion

BMI

body mass index

CARDIA

Coronary Artery Risk Development in Young Adults

CoV

coefficient of variation

CRP

C-reactive protein level

TIA

transient ischemic attack

Appendix. Authors

Name	Location	Contribution
Amber L. Bahorik, PhD	Department of Psychiatry and Behavioral Sciences, University of California, San Francisco	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Tina D. Hoang, MSPH	Northern California Institute Research for Research and Education, San Francisco, CA	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
David R. Jacobs, PhD	School of Public Health, University of Minnesota, Minneapolis	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Deborah A. Levine, MD, MPH	Department of Internal Medicine, and Department of Neurology, University of Michigan, Ann Arbor	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Kristine Yaffe, MD	Department of Psychiatry and Behavioral Sciences, Department of Neurology, and Department of Epidemiology and Biostatistics, University of California, San Francisco	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data; additional contributions: obtained funding

Study Funding

The Coronary Artery Risk Development in Young Adults (CARDIA) Study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with the University of Alabama at Birmingham (HHSN268201800005I and HHSN268201800007I), Northwestern University (HHSN268201800003I), University of Minnesota (HHSN268201800006I), and Kaiser Foundation Research Institute (HHSN268201800004I). The CARDIA Cognitive ancillary is supported by the National Institute on Aging (NIA) R01 AG063887. This work was also supported by NIA R35AG071916 (K.Y.). This manuscript has been reviewed by CARDIA for scientific content. The content is solely the responsibility of the authors and does not necessarily represent the views of the National Heart, Lung, and Blood Institute, the NIH, or the US Department of Health and Human Services.

Disclosure

K. Yaffe has served on the data safety monitoring board for Eli Lilly and several NIA-sponsored studies, has been a consultant for Alpha Cognition, has served on the board of directors for Alector Inc., has provided data and safety and monitoring services for the Dominantly Inherited Alzheimer Network Trails Unit, and has served on the Beeson Scientific Advisory Board and the Global Council on Brain Health. The other authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.