The field of cognitive aging is increasingly concerned with identifying early cognitive decline and the transition point from normal aging to the Alzheimer?s Disease (AD) continuum. Longitudinal assessments are necessary to monitor cognitive change. However, studies that involve repeated testing are subject to practice effects, which are typically defined as improvements in scores because of prior test exposure. Practice effects are important for studies of aging because they inflate performance, thereby obscuring the true degree of age- related cognitive decline expected at mid- and late life. If uncorrected for practice effects, stable performance in a longitudinal study may indicate cognitive decline that would go undetected based on typical norm-based classifications of impairment. Although cognitive decline is a likely a continuous process, cut points for impairment are necessary for determining when to alter patient care and when to enroll subjects in a study. Cut points for cognitive impairment, like cut points for biomarkers, are set because individuals with that level of performance are more likely to have other symptoms or a greater likelihood for disease progression. The misclassification of cognitive change may also obscure the relationship between cognition and biomarkers or risk factors for AD. Nevertheless, researchers almost always utilize uncorrected data, rely on purely statistical methods of practice effect correction, or simply covary for the number of visits. To directly address practice effects across two timepoints, a better method is to include replacement subjects who are naive to the tests, but age- and demographically-matched to returnees. Using this method, the Vietnam Era Twin Study of Aging (VETSA), demonstrated practice effects after six years, even when mean performance declined with age. Moreover, practice effect correction doubled the percentage of mild cognitive impairment (MCI) diagnoses while reducing the number of participants who reverted to normal. In this proposal I will extend this approach to practice effect correction by, for the first time, applying it across more than two assessments. Data will be from the VETSA and the Alzheimer?s Disease Neuroimaging Initiative (ADNI), which differ in participant age, retest interval, biomarkers, and number of assessments. I now have pilot data on the identification of ?pseudo? replacement subjects in ADNI. The method will be developed within ANDI and cross-validated in VETSA, which recruited ?true? replacement subjects. I hypothesize that accounting for practice effects will lead to earlier diagnoses of MCI, and a stronger signal between cognitive performance and biomarkers. With earlier detection of MCI, researchers and clinicians will be better able to track AD progression and monitor the effectiveness of potential treatments. Beyond the current proposal, a longer-term goal is to develop normative practice effect data (e.g., with NIH toolbox). By promoting earlier detection of cognitive decline and progression to AD-related disorders, normative practice effect data would have substantial nationwide public health implications.
Longitudinal studies of cognition are necessary for determining deterioration to Alzheimer?s-related disorders, and for fully understanding associations between early cognitive change and Alzheimer?s biomarkers or risk factors. Practice effects are apparent in a jump in later scores on repeated cognitive assessments, which may obscure progression to Alzheimer?s-related disorders and the relationship between cognition and biomarkers or risk factors. This proposal aims to develop a practice effect correction method for use when there are more than two assessments and an expected decline in cognitive ability over time.
