Please activate JavaScript!
Please install Adobe Flash Player, click here for download


86Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust Thursday 28 August 2014 – 9:00-12:30 Minisymposia M1 Statistical challenges in the epidemiology of aging Organizers: Carole Dufouil and Karen Leffondré M1.1 Methodological challenges in the epidemiology of aging from a reproducible research perspective SM Hofer1 1 University of Victoria, Victoria, Canada   The analysis of longitudinal observational data can take many forms and requires many decisions, with research findings and conclusions often found to differ across independent longitudinal studies addressing the same question. Differences in measurements, sample composition (e.g., age, cohort, country/culture), and statistical models (e.g., change/time function, covariate set, centering, treatment of incomplete data) can af- fect the replicability of results. The central aim of the MELODEM Initiative and the Integrative Analysis of Longitudinal Studies of Aging (IALSA) re- search network (NIH/NIA P01AG043362) is to optimize opportunities for replication and cross-validation across heterogeneous sources of longi- tudinal data by evaluating comparable conceptual and statistical models at the construct-level. I will provide an overview of the methodological challenges associated with comparative longitudinal research, including the comparability of alternative models of change, measurement harmo- nization and construct-level comparison, retest effects, distinguishing and contrasting between-person and within-person effects across studies, and evaluation of alternative models for change over time. These meth- odological challenges will be discussed within the context of reproducible research on aging-related outcomes. M1.2 Survival analysis aspects of the epidemiology of ageing N Keiding1 1 University of Copenhagen, Copenhagen, Denmark   Measurement of indicators of health and abilities is often restricted to ob- servations at long intervals, and for the elderly necessarily further com- plicated by the likely possibility that the individuals die. There is a lively debate on the best ways to define targets of inference and associated methods of analysis under this truncation by death sampling pattern. This talk will outline the positive body of concepts and tools from within survival and event history analysis which may be helpful in meeting these challenges, focusing on competing risks analysis with its descriptors cause-specific hazard rates and cumulative incidences, and on illness- death models, where an important issue is that of time origin: is survival measured from birth and/or first occurrence of disease? Disease preva- lence is naturally formalized in illness-death models. Since disease status is often recorded intermittently, estimation may need to handle interval- censored data. A main theme is to be precise about the study base, that is, the population about which one wants to make statements.   M1.3 Modelling issues in the longitudinal study of cognitive aging C Proust-Lima1,2 1 INSERM U897, Bordeaux, France, 2 University of Bordeaux, ISPED, Bordeaux, France   In cognitive aging studies, there is a growing interest in the description of change over time of cognitive functions and the evaluation of risk factors of cognitive change. Indeed, as dementia is characterized by a progressive and continuous decline of cognitive functions, these longitudinal analyses better capture the dynamics of disease progression than survival analyses describing time-to-progression. However, the study of cognitive change entails several modeling issues. First, markers of progression are psychometric tests with usually limited metric properties (ceiling/floor effects, curvilinearity) that translate in asymmetric distributions. To avoid large biases and misleading conclu- sions induced by these properties, mixed models adapted to psychomet- ric data can be preferred to the more standard linear mixed model. Second, multiple psychometric tests are usually collected and the inter- est is not directly in the change of one specific test but in the change of the underlying cognition that generated them. Latent process approaches that focus on the dynamics of a latent trait underlying a set of longitudi- nal outcomes are designed to address such multivariate and longitudinal data. Finally, cognitive change may be associated with clinical events, mainly dementia and death. These informative events potentially bias the cogni- tive change estimates when not taken into account and bring essential information for understanding the natural history of cognitive aging. Joint models of longitudinal outcomes and times-to-events account for this in- formative dropout. In this presentation, these statistical issues are gradually addressed and illustrated using data of a large prospective cohort study (PAQUID) on ce- rebral aging with a 22-year follow-up.   M1.4 Medical and conceptual challenges in conducting studies of the elderly MD Koeller1 1 Medical University of Vienna, Vienna, Austria   The objective of this talk is to identify conceptual challenges in medical research with older persons that should be addressed in the future work of biostatisticians cooperating and supporting geriatricians. Due to the diversity of medical problems and conditions in elderly patients clinical research studies require a specific design and analytic approach. These requirements are based on the multifactorial etiologies of geriatric syn- dromes and multimorbidity at the end of life. Moreover, manifold inter- ventions, multiple outcomes, ceiling effects, missing data, or different methods are factors which have to be taken into account in planning stud- ies in geriatric cohorts. Study design of multicomponent interventions in older persons is complicated. With respect to medical risk factors, partici- pants often cannot be randomly assigned to all possible interventions. Under ideal study conditions, all the participants are retained with com- plete data. But in geriatric patients functional assessments, for instance, are frequently limited and inappropriate, causing semicontinuous data and consequently floor or ceiling effects may result. Coincidences, bias, or ambiguity are potential threats to successful interpretation of results with good and clear conclusions. Therefore, interdisciplinary thinking and the cooperation of researchers from gerontology and geriatrics, as well as bi- ology, and the statistical discipline will have to be enhanced in the future.   ISCB 2014 Vienna, Austria • Minisymposia Thursday, 28th August 2014 • 9:00-12:30

Pages Overview