Abstract
Degree of memory impairment is often used to infer early versus late amnestic mild cognitive impairment (aMCI). Previously, 318 Alzheimer's Disease Neuroimaging Initiative participants with aMCI-determined by a single memory test-were divided based on Rey Auditory Verbal Learning Task (AVLT) delayed recall: AVLT-impaired (n = 225) and AVLT-normal (n = 93). Equally consistent with differential progression or differential diagnosis, the AVLT-impaired group had more abnormal Alzheimer's disease (AD) biomarkers, more neurodegeneration over time, and was more likely to develop AD. In the present study, higher AD polygenic risk scores were associated with increased odds of being AVLT-impaired (odds ratio 1.8, p < 0.001). Thus, impairment severity does not necessarily reflect early versus late aMCI because disease progression cannot alter polygenic risk. Presumed early MCI is likely a heterogeneous category that includes excess false-positives. The additional cognitive test improved diagnostic precision by reducing false positives. Impairment severity may reflect differences in underlying disease risk but cannot be used to infer early versus late MCI based on cross-sectional data alone.