Change in the self CFI scores by APOE ε4 carrier status (A) and the CDR-G progression status (B), change in the partner CFI scores by APOE ε4 carrier status (C) and the CDR-G progression status (D), and change in the self plus partner CFI scores by APOE ε4 carrier status (E) and the CDR-G progression status (F). Change from baseline is estimated by mixed models of repeated measures with baseline performance, age, educational level, and Geriatric Depression Scale score as covariates. The model treats time from baseline as a categorical variable and assumes compound symmetric correlation structure and heterogeneous variance over time. The shaded region marks the 95% CIs. The APOE group differences are significant (P < .05) for the partner CFI at months 36 and 48 and for the self plus partner CFI at months 24 to 48. The CDR-G progression group differences are significant at every visit except at months 3 and 12 for the self CFI.
Pearson product moment correlation coefficients are plotted over time, by version of the CFI (self, partner, and self plus partner), with 95% CIs (error bars). Signs for the correlations are inverted.
Amariglio RE, Donohue MC, Marshall GA, Rentz DM, Salmon DP, Ferris SH, Karantzoulis S, Aisen PS, Sperling RA, for the Alzheimer’s Disease Cooperative Study. Tracking Early Decline in Cognitive Function in Older Individuals at Risk for Alzheimer Disease DementiaThe Alzheimer’s Disease Cooperative Study Cognitive Function Instrument. JAMA Neurol. 2015;72(4):446-454. doi:10.1001/jamaneurol.2014.3375
Several large-scale Alzheimer disease (AD) secondary prevention trials have begun to target individuals at the preclinical stage. The success of these trials depends on validated outcome measures that are sensitive to early clinical progression in individuals who are initially asymptomatic.
To investigate the utility of the Cognitive Function Instrument (CFI) to track early changes in cognitive function in older individuals without clinical impairment at baseline.
Design, Setting, and Participants
Longitudinal study from February 2002 through February 2007 at participating Alzheimer’s Disease Cooperative Study sites. Individuals were followed up annually for 48 months after the baseline visit. The study included 468 healthy older individuals (Clinical Dementia Rating scale [CDR] global scores of 0, above cutoff on the modified Mini-Mental State Examination and Free and Cued Selective Reminding Test) (mean [SD] age, 79.4 [3.6] years; age range, 75.0-93.8 years). All study participants and their study partners completed the self and partner CFIs annually. Individuals also underwent concurrent annual neuropsychological assessment and APOE genotyping.
Main Outcomes and Measures
The CFI scores between clinical progressors (CDR score, ≥0.5) and nonprogressors (CDR score, 0) and between APOE ε4 carriers and noncarriers were compared. Correlations of change between the CFI scores and neuropsychological performance were assessed longitudinally.
At 48 months, group differences between clinical progressors and non-progressors were significant for self (2.13, SE=0.45, P < .001), partner (5.08, SE=0.59, P < .001), and self plus partner (7.04, SE=0.83, P < .001) CFI total scores. At month 48, APOE ε4 carriers had greater progression than noncarriers on the partner (1.10, SE=0.44, P < .012) and self plus partner (1.56, SE=0.63, P < .014) CFI scores. Both self and partner CFI change were associated with longitudinal cognitive decline (self, ρ=0.32, 95% CI, 0.13 to 0.46; partner, ρ=0.56, 95% CI, 0.42 to 0.68), although findings suggest self-report may be more accurate early in the process, whereas accuracy of partner report improves when there is progression to cognitive impairment.
Conclusions and Relevance
Demonstrating long-term clinical benefit will be critical for the success of recently launched secondary prevention trials. The CFI appears to be a brief, but informative potential outcome measure that provides insight into functional abilities at the earliest stages of disease.
Previously published guidelines have outlined a preclinical phase of Alzheimer disease (AD) in which individuals are still clinically normal but may have subtle evidence of early cognitive change in the context of amyloidosis and neuronal injury.1 This phase provides the opportunity to intervene at earlier stages of disease than was previously possible.2 Guidance from the Food and Drug Administration states that a primary cognitive outcome measure may suffice for provisional approval of a preclinical treatment but eventually would need to be supported by evidence of long-term functional benefit.3 Currently, sensitive cognitive measures are available for secondary prevention trials,4,5 but a companion functional measure has yet to be fully realized.6
Subjective report of cognitive functioning in everyday life is increasingly thought to be a sensitive indicator of decline, even at the preclinical stages of AD.7- 13 During the last decade, the Alzheimer’s Disease Cooperative Study (ADCS) has developed the Cognitive Function Instrument (CFI), which is intended to detect early changes in cognitive and functional abilities in individuals without clinical impairment.14 The CFI includes 14 questions that are asked of the participant and a study partner separately that efficiently probe the full realm of subjective cognitive concerns found in older adults.15 Unlike functional outcomes typically used in clinical trials at the stage of mild cognitive impairment (MCI) or dementia, the CFI does not require an in-person interview or clinician judgment and measures both participant and study partner report, the combination of which is sensitive at the earliest stages of disease.16
In the current study, the goal was to determine whether the CFI is a sensitive measure in tracking longitudinal change in cognitive function in older individuals without cognitive impairment at baseline. During 4 years, we assessed the CFI’s (self and partner) association with clinical progression (on the Clinical Dementia Rating scale [CDR]), APOE ε4 carrier status, and performance on a cognitive composite. We ultimately sought to establish the CFI as a useful functional measure appropriate for prevention trials.
A total of 468 older individuals (60.3% female) with a mean (SD) age of 79.4 (3.6) years (age range, 75.0-93.8 years) participated in the study. Participants met inclusion criteria if they were in good physical and mental health, had no significant medical illnesses that would interfere with participation (eg, active malignant tumor or stroke), and no exclusionary medications (eg, antipsychotic agents). Participants were required to have a qualified study partner who was willing to provide information about their daily function and who had contact with the participant at least 2 times per week. The participants were enrolled at multiple ADCS sites. The local institutional review boards at each site approved the study protocol and informed consent form before the initiation of participant recruitment. All participants provided written informed consent after the procedures of the study had been fully explained.
In the current sample, participants had a CDR global (CDR-G)17 score of 0 at baseline, a modified Mini-Mental State Examination (3MSE) score18 of at least 88 for participants with greater than 8 years of education and at least 80 for participants with less education (taken from the Women’s Health Initiative Memory Study), and a total score greater than 44 on the Free and Cued Selective Reminding Test (FCSRT), which includes both free and cued recall.19 The CFI was not used to determine participant eligibility in this study. Individuals were followed up annually for 48 months after the baseline visit.
Participants and study partners were mailed the CFI 4 weeks before each annual assessment and were asked to return the questionnaire by mail. Participants and study partners were asked to complete the CFI independently. The study partner was not allowed to consult the participant but could consult anyone else. The CFI contains 14 questions (Table 1 and Table 2): one version for the participant and one version for the study partner with the same questions. Questions were originally derived from common probes across clinical assessments of aging and dementia14 and include items regarding memory decline (eg, compared to a year ago, memory has declined), appraisal of cognitive difficulties (eg, misplacing belongings more often), and functional abilities (eg, need more help remembering appointments). Responses were coded as 1 for yes, 0 for no, and 0.5 for maybe and were summed to create a total score.
Objective cognitive performance was assessed annually with a cognitive battery called the modified Alzheimer’s Disease Cooperative Study Preclinical Alzheimer Cognitive Composite (mADCS-PACC), a modified version of the ADCS-PACC,4 described in a separate study, which has demonstrated sensitivity in detecting cognitive decline at the preclinical stage of AD. Briefly, the mADCS-PACC consisted of (1) total recall from the FCSRT (0-48 words),19 (2) New York University Paragraph Recall,20 (3) Digit-Symbol Substitution Test score from the Wechsler Adult Intelligence Scale–Revised (0-93 symbols),21 and (4) 3MSE18 score. The New York University Paragraph Recall and the Digit Symbol Substitution Test were not given at baseline but at all subsequent visits.
At each annual follow-up visit, participants were screened for the development of MCI or dementia based the same cutoffs used as inclusion criteria (3MSE and FCSRT). If an individual fell below either 2 cutoffs on the cognitive evaluation, then the site physician initiated a clinical consensus to determine whether the participant met Petersen criteria22 for MCI or Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) for dementia and, if so, whether the probable cause was AD.
Among those with at least one follow-up CFI observation, we summarized means (SDs) and numbers (percentages) of characteristics at baseline by groups of interest. One group of interest included individuals whose scores progressed from 0 to 0.5 or higher on the CDR-G at any time point after baseline. None of the study participants who progressed on the CDR regressed on the CDR at a later time point. We also compared individuals who had at least one APOE ε4 allele (APOE ε4 positive) against noncarriers (APOE ε4 negative). Groups were compared using the Pearson χ2 test for categorical data and the 2-sample t test for continuous data. We assessed internal consistency, using the Cronbach α, on the items of the CFI in our sample for the self and study partner versions. In individuals who were CDR-G stable and APOE ε4 negative, we estimated intraclass correlation coefficients to assess test-retest reliability from baseline to 12 months later.23
To assess group differences in longitudinal change, we applied a mixed model of repeated measures24 with baseline CFI score, baseline Geriatric Depression Scale score,25 age, educational level, and race considered as covariates. Time was treated as a categorical variable. We assumed a compound symmetric correlation structure and heterogeneous variance over time. No adjustments were made for multiple comparisons. Secondary analyses compared CFI items that represented appraisal of cognitive abilities in everyday life (eg, repeating questions) to those that were more related to functional abilities (eg, change in ability to use appliances) in predicting outcomes. We also regressed CDR-G progression status on the CFI and PACC baseline scores to assess the predictive value of each predictor separately and together. The predictive values of these logistic regression models were compared using the Akaike Information Criterion (AIC).26 This analysis was repeated with the CFI and PACC change scores, estimated by the slope estimates from participant-level ordinary least squares regression.
To assess the correlation between the CFI versions (self, partner, and self plus partner) and the mADCS-PACC, we estimated cross-sectional Pearson product moment correlation coefficients at each visit. We used the nonparametric bootstrap, resampling participants with replacement 10 000 times, to obtain CIs for the pairwise differences between these correlations at each time point (self vs partner, self vs self plus partner, and partner vs self plus partner). We also applied a multivariate outcome linear mixed-effect model approach to estimate the correlation of change in the CFI and mADCS-PACC.27 The CIs were again derived by nonparametric bootstrap in which we resampled participants, with replacement, 1000 times and refit the mixed-effect model for each resample.
Table 3 and Table 4 provide the baseline characteristics by the groups of interest. The CDR-G stable group had better performance than the CDR-G progressor group on the FCSRT free recall subscale, 3MSE, and all 3 versions of the CFI (self, partner, and self plus partner). The APOE ε4–negative group was older and had higher Geriatric Depression Scale scores at baseline than the APOE ε4–positive group. Individuals who dropped out of the study were more likely to be older, to be of minority status, and to have lower educational levels. Performance on the 3MSE and the PACC were lower at the time of the last visit, but no differences were found on the CDR.
The Cronbach α at baseline was 0.78 for the self CFI and 0.85 for the partner CFI. In individuals who were CDR-G stable and APOE ε4 noncarriers, intraclass correlations were 0.73 for the self CFI and 0.54 for the partner CFI (5.0% of the study partners changed from baseline to month 12).
Figure 1 shows the estimated group differences in CFI change from baseline by APOE and CDR-G progressor groups. When compared with the APOE ε4–negative participants, APOE ε4–positive participants have worse functioning on the CFI over time. The mean (SE) APOE group differences for the partner CFI were 1.04 (0.38) at 36 months (P = .007) and 1.10 (0.44) at 48 months (P = .01), and the mean (SE) APOE group differences for the self plus partner CFI were 0.94 (0.40) at 24 months (P = .02), 1.42 (0.53) at 36 months (P = .007), and 1.56 (0.63) at 48 months (P = .01). When compared with the CDR-G stable participants, CDR-G progressors have worse functioning on the CFI over time. The CDR-G progression group differences were significant for the self, partner, and self plus partner CFIs at every visit except at months 3 and 12 for the self CFI. At month 48, the mean (SE) group differences were 2.13 (0.45) for the self CFI (P < .001), 5.08 (0.59) for the partner CFI (P < .001), and 7.04 (0.83) for the self plus partner CFI (P < .001). A composite of items grouped by functional abilities (eg, remembering appointments) vs a composite of metacognitive items (eg, repeating conversations) performed similarly at separating CDR and APOE groups longitudinally.
Figure 2 shows the cross-sectional correlation between the CFI and the mADCS-PACC. Worse functioning on the CFI was associated with worse cognition on the mADCS-PACC. The correlation between self CFI and the mADCS-PACC increased from 0.20 (95% CI, 0.12-0.29) at baseline to 0.39 (95% CI, 0.30-0.48) at month 48, whereas the correlation between the partner CFI and the mADCS-PACC increased from 0.09 (95% CI, 0.002-0.18) at baseline to 0.44 (95% CI, 0.35-0.52) at month 48. The correlation between the self plus partner CFI and the mADCS-PACC increased from 0.18 (95% CI, 0.10- 0.27) at baseline to 0.49 (95% CI, 0.40-0.57) at month 48. The bootstrap 95% CIs for the pairwise differences between correlations with the mADCS-PACC excluded 0 for all differences except between the self and self plus partner CFIs at baseline and month 24 and the self and partner CFIs at months 12, 36, and 48. The multivariate mixed-effect model analysis found that the correlations of change were 0.32 (95% CI, 0.13-0.46) between the self CFI and the mADCS-PACC, 0.56 (95% CI, 0.42-0.68) between the partner CFI and the mADCS-PACC, and 0.58 (95% CI, 0.44-0.70) between the self plus partner CFI and the mADCS-PACC. The correlation of change between the mADCS-PACC and the self plus partner CFI was significantly stronger than with the self CFI (0.27; 95% CI, 0.09- 0.46). The correlation of change between the mADCS-PACC and the self plus partner CFI was stronger than with the partner CFI (0.02; 95% CI, −0.10 to 0.13) and stronger with the partner CFI than with the self CFI (0.25; 95% CI, −0.02 to 0.53), but these differences were not significant at the .05 level.
The logistic regression models of CDR-G progression found that the baseline mADCS-PACC and CFI scores were independently predictive. Each point increase on the PACC was associated with a decreased risk of progression (odds ratio, 0.965; 95% CI, 0.95-0.98; P < .001), and each point increase on the CFI was associated with increased risk of progression (odds ratio, 1.013; 95% CI, 1.01-1.02; P < .001). The model that included both measures had the best predictive value (AIC = −100.2), followed by the models with the mADCS-PACC alone (AIC = −89.3) and the CFI alone (AIC = −78.1).
Overall, we found that self and partner report of change in cognitive function on the CFI was associated with traditional measures of clinical progression during 4 years. When comparing individuals who had clinical progression (CDR-G score, >0) with those who remained stable (CDR-G score, 0), a significant separation was found between groups, such that CDR-G progressors exhibited higher CFI scores and a greater increase in the CFI scores over time than did the CDR-G stable individuals. These findings held for partner report and self-report. The combination of the self and partner CFI revealed a slight advantage over individual report, suggesting that both perspectives on decline might be valuable during a 4-year observational period.
The CFI remained a predictor of CDR progression when objective cognitive performance was also added as a predictor, suggesting that it independently contributes to longitudinal outcomes but that the combination of both measures may be particularly predictive. Findings are in support of previous work that found that subjective and objective measures improve predictive ability in individuals without clinical impairment7 and with MCI.28
When comparing APOE ε4 carriers and noncarriers, a significant separation was found between carrier groups by month 36 for the partner CFI, whereas the self CFI was not different between carrier groups at any time point. However, the combination of the self and partner CFI differentiated between groups at month 24 and revealed the greatest separation between groups, although not statistically different from self and study partner alone.
Additional analyses did not reveal significant differences between the CFI items that represented appraisal of cognitive abilities in everyday life (eg, repeating questions) compared with report of functional abilities (eg, change in ability to use appliances). Questions regarding cognitive difficulties or functional abilities performed equally as well at differentiating between progressors and nonprogressors or APOE ε4 carriers and noncarriers, suggesting that both are valuable in assessing subjective report of everyday functioning.
When assessing the correlation between longitudinal change on the CFI compared with longitudinal change on an objective cognitive composite, we found that increased partner and self CFI scores were associated with greater objective cognitive decline. When we examined the correlation between the CFI and objective cognitive performance on the mADCS-PACC at each time point, we saw, at baseline and month 24, that the self CFI revealed significantly stronger correlations with cognitive performance than the partner CFI. Self-reports and partner reports were significantly inferior to the combined report at months 12, 36, and 48, suggesting that the combination is particularly powerful in detecting subtle cognitive decline. Self-report was significantly superior to partner report at baseline and month 24, whereas partner report was numerically superior (but did not reach statistical significance) at months 36 and 48. One interpretation is that although self-report is more reliably correlated with cognition earlier in the process of decline, partner report might become more useful later with development of anosognosia.16,29
Reliability analyses revealed that internal consistency of items was at acceptable levels for self-reports and partner reports. Intraclass correlations revealed that in individuals who were CDR-G stable and APOE ε4 noncarriers, self-report was more reliable after 12 months than for the study partner despite only a 5% turnover rate in study partners. This finding may be because study partners are not in a caregiving role because participants were all independent in their everyday activities at baseline and suggests that the combination of self-report and partner report may be more reliable in tracking change in individuals who begin at an asymptomatic stage.
Results suggest that the CFI can serve as a sensitive functional outcome measure in secondary prevention trials. Brief and easily administered on an annual basis in the current study, it contains questions that cover the full realm of early functional change. Of importance, the CFI can be self-administered at home, with forms mailed,14 or administered over the telephone or transmitted electronically; thus, the CFI could be used in large, lengthy prevention trials with minimal in-person contact between participants and investigators. Historically, clinical trials that involve patients with MCI or AD dementia have included more detailed functional measures, such as the CDR,17 which requires a physician’s judgment, or the Disability Assessment for Dementia,30 which is weighted heavily toward a study partner’s report. In addition to the CFI, the ADCS has developed a more extensive functional measure, called the Activities of Daily Living–Prevention Instrument.31 This questionnaire also has sensitivity in detecting early functional change; however, it takes longer to administer than the CFI and does not include questions about self-appraisal of memory function, a feature that may be important to capture early progression along the preclinical stages of AD.
Several potential limitations of this study deserve mention. Biomarker data were not available to confirm the cause of cognitive decline in our sample. Although items of the CFI were originally selected to target changes commonly experienced in functional impairment due to AD, it is possible that this instrument is sensitive to changes associated with other causes. Follow-up studies that explore associations with putative AD biomarkers will be informative. In addition, there is some overlap in the types of questions asked in the CFI and the CDR, which may lead to enhancement of findings when looking at clinical progression outcomes. However, differences on CFI were found at baseline when all participants had a CDR score of 0, suggesting that the CFI could be more sensitive than the CDR in a sample without clinical impairment. Furthermore, we found that the CFI tracked with objective cognitive decline, providing support for its utility.
As the AD field moves toward prevention at the preclinical stages of disease, many new hurdles emerge. In addition to identifying the appropriate target for disease modification, finding the right tools to detect the earliest evidence of clinical progression is challenging. Demonstrating long-term clinical benefit will be critical because maintaining independence in everyday functioning is what matters most to patients and their families. Subjective assessment of an individual’s level of functioning over time with the CFI may prove to be a sensitive and efficient outcome for secondary prevention trials in preclinical AD.
Accepted for Publication: September 18, 2014.
Corresponding Author: Rebecca E. Amariglio, PhD, Center for Alzheimer Research and Treatment, Department of Neurology, Brigham and Women’s Hospital, 221 Longwood Ave, Boston, MA 02115 (firstname.lastname@example.org).
Published Online: February 23, 2015. doi:10.1001/jamaneurol.2014.3375.
Author Contributions: Dr Amariglio had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Rentz, Salmon, Ferris, Aisen, Sperling.
Acquisition, analysis, or interpretation of data: Amariglio, Donohue, Marshall, Ferris, Karantzoulis, Aisen, Sperling.
Drafting of the manuscript: Amariglio, Donohue.
Critical revision of the manuscript for important intellectual content: Donohue, Marshall, Rentz, Salmon, Ferris, Karantzoulis, Aisen, Sperling.
Statistical analysis: Amariglio, Donohue, Salmon.
Obtained funding: Aisen.
Administrative, technical, or material support: Aisen.
Study supervision: Rentz, Salmon, Aisen, Sperling.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by grant NIRG-12-243012 from the Alzheimer’s Association (Dr Amariglio) and grants U19 AG10483 (Dr Aisen), K24 AG035007 (Dr Sperling), P01 AG036694 (Dr Sperling), and K23AG044431 (Dr Amariglio) from the National Institutes of Health.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Alzheimer’s Disease Cooperative Study Group: Oregon Health Sciences University, Portland (Kathy Wild, PhD, and Georgene Siemjen, MS, RN); University of Southern California, Los Angeles (Lon Schneider, MD, and Alla Sverdlik); University of California at San Diego, La Jolla (Leon Thal, MD, and Mary Pay, RN, NP); Michigan Alzheimer's Disease Research Center, Ann Arbor (R. Scott Turner, MD, PhD, and Linda V. Nyquist, PhD); Mayo Clinic, Rochester MN (Ronald C. Petersen, PhD, MD, and Joan McCormick, BSN); Columbia University, New York, NY (Karen Bell, MD, and Ruth Tejeda, MD); Washington University, St Louis, MO (John Morris, MD, and Pamela Millsap, MSN); University of Alabama at Birmingham, Birmingham (Lindy Harrell, MD, PhD, and Jo Ann Parrish, LPN); Mt Sinai Medical Center, New York, NY (Karen Dahlman, PhD, and Adriana DiMatteo, MA); Rush Alzheimer's Disease Center, Chicago, IL (Neelum T. Aggaewal, MD, and Rose Marie Ferraro, LVN, AS); Wien Center for Memory Disorders, Mount Sinai Medical Center, Miami Beach, FL (Ranjan Duara, MD, and Peggy D. Roberts); University Memory and Aging Center, Cleveland, OH (Paula Ogrocki, PhD, and Nancy A. Slocum, RN, MPH); Johns Hopkins University, Baltimore, MD (Cynthia Munro, PhD); University of South Florida, Tampa, FL (Eric Pfeiffer, MD, and Dorothy Baxter, MPH); New York University School of Medicine, New York, NY (Steven Ferris, PhD, and Nicole Martingano, BA); University of Pennsylvania, Philadelphia (Christopher Clark, MD, and Kris Gravanda, BA); University of Pittsburgh, Pittsburgh, PA (Judith Saxton, PhD, and Patrick Ketchel, MEd); University of Rochester, Rochester, NY (J. Michael Ryan, MD, and Colleen McCallim, MSW); University of California at Irvine, Irvine (Ruth Mulnard, DNSc, and Catherine McAdams-Ortiz, MSN); University of Texas Southwestern Medical Center, Dallas (Myron Weiner, MD, and Robbin Peck, AD); Emory University, Atlanta, GA (Felicia Goldstein, PhD, and Lisa Kilpatrick, MS); University of California at Los Angeles, Los Angeles (John Ringman, MD, and Susan O'Connor, RNC); Mayo Clinic, Jacksonville, FL (Neill Graff-Radford, MD, and Francine Parfitt, MSH, CCRC); Indiana University Alzheimer's Center, Indianapolis (Martin Farlow, MD, and Nicki Coleman, RN); Memorial Hospital of Rhode Island (Brown), Pawtucket (Brian Ott, MD, and Michael Pimental, CMA); Yale University, New Haven, CT (Christopher van Dyck, MD, and Martha MacAvoy, PhD); University of California at Davis, Sacramento (Charles DeCarli, MD, and Bobbi Henk, RN, MSN); Baumel-Eisner, Boca Raton, FL (Larry Eisner, MD, and Rebecca Radzivill, MS); Baumel-Eisner, Ft Lauderdale, FL (Barry Baumel, MD, and Magdalena Szymczak); Baumel-Eisner, Miami Beach, FL (Beth Safirstein, MD, and Melissa Perez-Velasco, BS); University of Nevada at Las Vegas, Las Vegas (Charles Bernick, MD, and Marie L. Stallbaum, BSN); Northwestern University, Chicago, IL (M. Marsel Mesulam, MD, and Laura Herzog, MA); The Medical University of South Carolina, North Charleston (Jacobo Mintzer, MD, and Effie Hatchett, RNC); Georgetown University, Washington, DC (Paul S. Aisen, MD, and Carolyn Ward, MSPH); Brigham and Women's Hospital, Boston, MA (Reisa Sperling, MD, MMSc, and Kara Campoba, MA, PA); Stanford/Veterans Affairs Aging Clinical Research Center, Palo Alto, CA (Joy Taylor, PhD, and Heather Fiedler-Greene, MA); Sun Health Research Institute, Sun City, AZ (Donald Connor, PhD, and Suhair Stipho, MB, CHB); Boston University School of Medicine, Boston, MA (Robert C. Green, MD, PhD, and Mary-Tara Roth, RN, MSN, MPH).
Additional Contributions: We acknowledge the staff of the ADCS, the site personnel from the ADCS–Prevention Instrument Trial, and all the participants who dedicated their time to this study.
Correction: This article was corrected on March 9, 2015, to fix an error in Figure 1.