SA 1: To complete the development of an interactive website and searchable data base for enhancing coordinated, collaborative analyses of longitudinal studies. Improvements have been made to both the infrastructure and content of the IALSA website used for planning, managing, and communicating results of the related scientific projects. We have transferred the site (www.ialsa.org) to a Drupal-based system that is locally supported by UVic technical support. In addition to the structural/administrative development of the study comparison utility, cognitive, psychosocial, and health-related constructs were further articulated and further content and detail regarding collected data were obtained from network members and uploaded to the newly programmed software, including a range of biospecimen data in collaboration with the Biolibrary for Aging Relevant Data (BARD).
SA 2: To document general patterns of population average and individual variation in change in cognitive capabilities across major longitudinal studies on aging.
Project B. This analysis project evaluates the effects of age, education, and sex on intraindividual trajectories in the MMSE across studies and compares Identity link (standard) and Tobit link growth models, which adjust for the expected ceiling with this screening measure. The comparability of the between-person (age difference) and within-person (time change) estimates is addressed. We have completed final analysis of six longitudinal studies and have reported the results in a single, comprehensive report of cross-study patterns and predictors of change in the MMSE. Initial age and years of education both predicted level at study entry. Age predicted rate of change in MMSE scores, with older individuals starting lower and declining faster. Education was a significant predictor of change in MMSE only for the OCTO-Twin study, where those with more education declined more slowly, however this association was no longer statistically significant once individuals diagnosed with dementia were omitted from the analysis. In general, results from the Tobit models were consistent with the identity link models, indicating that the impact of ceiling effects in these studies were not large. This project is completed, submitted, and awaiting review:
Piccinin, A. M., Muniz, G., Deary, I., Deeg, D., Johansson, B., MacKinnon, A. Reynolds, C.A., Spiro, A. III, Starr, J. M., Thorvaldsson, V. & Hofer, S. M. (submitted to The Journals of Gerontology: Psych Sciences). Integrative analysis of longitudinal studies on aging: Coordinated Analysis of age, sex, and education effects on change in MMSE scores.
Related collaboration on ESRC award: Bayesian methods for combining multiple individual and aggregate data sources in observational studies II, PI: S. Richardson & N. Best, Imperial College London. This project, funded by the ESRC (UK), has as one of its specific aims the development and application of new methods for Bayesian evidence synthesis related to cross-study analysis of IALSA projects B and C. A model parallel to that of the MMSE paper is being implemented, and extended to include all data in a single model that will address differences in sample characteristics across the studies. Separate and joint analyses were conducted for CLS, HOPE, LASA, Octo-Twin and SATSA. A manuscript has been completed, and is currently undergoing final review by the co-authors in order to be submitted.
Molitor, J.N., Molitor, J.T., Best, N., Richardson, S. Piccinin, A.M., Christensen, H., Starr, J., Deeg, D., Johansson, B., Reynolds, C., & Hofer, S.M. Does education affect cognitive reserve? Combining five international aging studies.
Project C.This project focuses on an evaluation of aggregate between-person (BP) and within-person (WP) age trends in distinct cognitive outcomes within and across studies, conditional on age, sex, and education, and evaluating whether differences across BP and WP effects can be plausibly explained by either retest (test exposure) or cohort effects (i.e., Flynn effect of increasing IQ scores). Previous reports of age-related change in cognition have focused on BP age difference trends, or a combination of both BP and WP effects and only rarely has the equality of BP and WP age trends (i.e., convergence) been tested. We have developed a general analysis protocol, completed preliminary analyses of measures of fluid reasoning and memory in
HOPE, LASA, and OCTO-Twin data, and are working to implement this analysis protocol in other IALSA-affiliated studies. To papers have been published that provide overviews and general analytical approaches used in this project.
Hoffman, L., Hofer, S. M., & Sliwinski, M. J. (in press). On the confounds among retest gains and age-cohort differences in the estimation of within-person change in longitudinal studies: A simulation study. Psychology and Aging.
Sliwinski, M. J., Hoffman, L., & Hofer, S. M. (2010). Evaluating convergence of within-person change and between-person age differences in age-heterogeneous longitudinal studies. Research in Human Development, 7, 45-60.
Project D: The aim of this project is to identify and describe the level and rate of change in MMSE at the individual and aggregate population level prior to diagnosis of dementia. The focus is on the impact of dementia on the longitudinal progression of changes in the MMSE. Many IALSA studies have independent diagnoses of dementia, and this will be used to examine changes in MMSE prior to diagnosis. This will also allow us to evaluate criteria for dementia cut-off scores based on both cross-sectional and longitudinal information, which will permit a basis upon which to make best use of the MMSE as proxy measure of dementia in longitudinal studies for which clinical diagnosis is not available. The focus is the impact of dementia on the progression of changes in the MMSE prior to diagnosis. Logistic regression and growth mixture models will allow the evaluation of criteria for dementia cut-off scores using longitudinal information from the MMSE and will provide an empirical basis to make use of the MMSE as proxy measure of dementia in studies for which clinical diagnosis is not available. Random change point and Tobit “censored outcome” models will again be applied in these analyses, which will be possible in data from CCS, CLS, H-70, HOPE, OBAS, OCTO-Twin, and SATSA. A specific effort will also be made to translate these findings into practical guidelines for clinical settings. This project is now underway, with an expected completion date of summer 2012.
Project E: Mortality-related patterns of cognitive change. One paper has been accepted for publication, with replication in other IALSA-affiliated studies to follow, and a second is under review.
A. M. Piccinin, G. Muniz, F. Matthews, B. Johansson (in press). Terminal decline from within and between person perspectives, accounting for incident dementia. Journal of Gerontology: Psych Sciences.
Terminal cognitive decline was modeled from a within-person perspective using a time-to-death time structure. WAIS Information and Block Design scores from 461 OCTO-Twin Study participants with confirmed death dates were modeled using quadratic growth curve models controlling for age and distance from death at study entry, sex, education and dementia diagnosis. Information scores showed evidence of within-person accelerating decline in the no dementia group, and stronger acceleration for incident dementia. Block Design scores declined, but did not accelerate within or across individuals. Strategies for focusing models on longitudinal aspects of available data are discussed.
G. Muniz Terrera, A van den Hout, A. M. Piccinin, S. M Hofer, F. E. Matthews & the CC75C Collaboration Group (submitted to RSS, Series A). Investigating terminal decline: results from a UK population-based study of ageing.
The terminal decline hypothesis states that in the proximity of death, individuals experience a change in rate of decline of their cognitive abilities. In this context, years to live are better indicators of cognitive change than chronological age. Using a time to death time metric to align individuals according to a more meaningful underlying process, we examined the terminal decline hypothesis fitting a random change point model that allowed us to identify individual-specific onset of terminal decline. We show that the onset of terminal decline occurs at about 6 years before death with differences in performance before death by gender, education and age at death and by gender, disability status at study entry and age at death in rate of change after the onset of the terminal decline process.
SA 3: To evaluate the commonality and specificity of cognitive changes with age, including tests of psychological theories of speed and working memory, from a within-person change perspective.
Project F: Evaluation of the speed mediation hypothesis from a within-person perspective. Aging-related decline in processing speed has long been considered a key driver of changes in other cognitive functions with age. While the majority of empirical evidence for the processing speed hypothesis has been obtained from cross-sectional analyses of age differences, studies of intraindividual change suggest an attenuated mediation effect of age via processing speed. This project examines the speed-mediation hypothesis, and describes the importance of alternative and complementary tests of within-person age change and variation using multiple longitudinal studies associated with the IALSA network. The main objective of the current project is to examine the importance of processing speed for explaining change in cognitive functioning following the approach used by Sliwinski and Buschke (1999; 2004) as well as extensions to this model that could further improve our understanding of the processing speed hypothesis and longitudinal analytic techniques. The analyses have been completed on OCTO-Twin and LASA data. Processing-speed was a significant predictor of cognitive functioning both cross-sectionally and longitudinally. The percent mediated effect of age on cognitive functioning via processing speed for the between-person and within-person fixed and random effects depended on the measure of cognition used. These results provide evidence for a moderate degree of interdependency among different types of cognitive functioning.
Robitaille, A., Muniz, G., Piccinin, A. M., IALSA Studies, & Hofer, S. M. (in preparation). Does processing speed account for aging-related change in other cognitive functions? An evaluation of between-person and within-person effects across IALSA studies.
Methodological issues related to Project F.As an adjunct to this project, therefore, acomparison was made of different approaches for the analysis of multivariate longitudinal change, including the parallel latent growth curve model, the directional bivariate latent growth curve model, and the latent growth curve model with a time-varying covariate. Processing speed and other cognitive function data from the OCTO-Twin and LASA studieswere used to illustrate the similarities and differences between the models. The three models indicated similar interrelationships between initial levels, latent change, and occasion-specific residuals for processing speed and other cognitive functions. However, differences in the parameter estimates exist between the models. The differences in parameter estimates between the models, the research questions that can be answered by each model and an overview of the advantages and disadvantages of each approach are discussed. Choosing one model over another should be given some thought and should take into account theoretical and empirical aspects of the research question. A manuscript is almost complete and will be submitted for publication in summer 2011.
Robitaille, A., Muniz, G., Piccinin, A. M., Aartsen, M., Johansson, B., & Hofer, S. M. (in preparation). Multivariate longitudinal modeling of cognitive change: Relationships between processing speed and other cognitive outcomes.
Related Research: Influence of health conditions on cognitive function in later life.This project has been undertaken by University of Victoria doctoral student, Catherine Sparks, and evaluates cognitive change in the context of change in health using different operational definitions and specifications of health (i.e., independent condition, comorbidity index), and compares patterns of results across multiple longitudinal studies. The aim of this study is to discriminate health-related cognitive change due to identified disease processes and to evaluate the impact of disease classifications on distinct cognitive outcomes. We have completed analysis of three longitudinal studies. Effects of disease were not consistently found across studies or across cognitive outcomes within each study. Hypertension, cardiovascular disease, and diabetes show the greatest impact across studies, with measures of processing speed most sensitive to changes in health. Differences are seen across studies in level of significance and magnitude of effects. In general, the OCTO-Twin study showed the greatest impact of change in health on level and rate of change in cognitive outcomes. In HRS, health conditions were consistently related to level, but less consistency for rate of change. The ELSA study showed little impact on level or rate of change. Having more than one condition had a significant impact on initial level, but not on rate of change, in the OCTO-Twin and HRS studies only.