Longitudinal standards for mid-life cognitive performance: Identifying abnormal within-person changes in the Wisconsin Registry for Alzheimer’s Prevention

A major challenge in the field of cognitive aging is differentiating disease-related cognitive change that has not yet become overt impairment from the more gradual decline in performance expected with normal aging. Published normative reference values are nearly always for single time point performances rather than longitudinal change in performance. To gain insight into how we might tackle the problem of identifying worrisome trajectories, we borrow a method from anthropometry: the development of standards that are conditional on an individuals past measurements. We use quantile regression to create growth-curve-like models of performance on several common neuropsychological tests of memory and executive function while accounting for age, sex, education, estimated verbal ability, and past performance on the test, and then use these to estimate individuals percentile ranks. Choosing the 7th percentile as a threshold (corresponding to approximately 1.5 standard deviations below the expected mean), we then explore relationships between subthreshold performance, clinical outcomes, and subjective impairment. Participants whose performance fell below the 7th percentile were more likely to be given an abnormal research diagnosis at the current visit, but not at later visits. Performance below this threshold was also linked to subjective and informant reports of worsening memory function. We discuss potential uses of this method in theoretical and applied research and clinical settings.