What is a normal change in memory? n=1 classification of clinically relevant cognitive decline

Abstract

Background: Clinicians often have to decide whether a person’s reported or measured change in cognition reflects normal aging or clinically-relevant cognitive decline. For most cognitive tests, there is no evidence-based guidance for how much change is ‘normal’ over a given timeframe. This study aimed to statistically determine the limits of clinically-significant memory change among healthy older adults and those with mild cognitive impairment (MCI). Methods: Longitudinal cognitive performance was examined in 21 healthy older adults and 81 individuals with clinically-defined MCI. Participants were assessed on Cantab Paired Associate Learning (PAL), a non-verbal episodic memory test, at baseline, 10 months and 20 months as part of a comprehensive neuropsychological assessment. They were classified as amnestic-MCI if there was a deficit (performance <10th percentile relative to norms) in memory tests (with our without impairment in other cognitive tests), and nonamnestic MCI if deficits were limited to non-memory domains. Over the three assessments, MCI was stable, resolved or progressed to dementia. The performance of control participants was used to determine the normal limits for initial and longitudinal change in performance, against which participants with stable or resolved MCI and those who progressed to dementia could be compared on an individual basis. Results: In healthy controls, PAL showed good stability using Bland-Altmann analysis. Na-MCI, and those with resolved MCI performed at a level comparable to healthy controls on each occasion. Patients with stable a-MCI or AD patients were significantly impaired relative to other groups. This persisted even when controlling for age, education, gender and baseline level of performance. Individual estimates of clinically significant impairment and change were able to reliably detect patients who would subsequently develop dementia, or had stable a-MCI. Conclusions: Repeated assessments with the Cantab PAL test are feasible, show good psychometric properties, and can characterise typical and atypical patterns of change in episodic memory. A repeat assessment of memory at <1 year can reliably inform clinical decisions about the presence or absence of pathological cognitive decline.