Serial Position Effect Profiles and Their Neuroanatomical Correlates: Predictors of Conversion to Alzheimer\u27s Disease

The current study was designed to determine whether targeted, premorbid, neuropsychological measures of the serial position effect (SPE) can detect and explain risk for later development of Alzheimer’s disease (AD). The study tested the utility of SPE measures in healthy controls (HC) and individuals already diagnosed with mild cognitive impairment (MCI) or AD. Aim 1 was to determine whether these sensitive, valid neuropsychological measures can explain disease risk. SPE of list-learning are highly sensitive cognitive markers that capture important elements of both linguistic and amnestic mechanisms of encoding, learning, and retrieval. Using the Rey Auditory Verbal Learning Test (RAVLT), we tested two measures of serial position effect scores (SPE-Index) calculated as accuracy recall at different serial positions at Learning, Short-Delay or Long-Delay, as well as SPE profile scores (SPE-Contrast), which compared accuracy recall of two SPE positions. The three SPE-contrast scores were calculated from primacy, middle, and recency list regions at Learning (Learning), short and long delay recall (SD, LD) trials as follows: (1) J-Shape captures difference between recency and primacy scores at Learning (RecencyLearning - PrimacyLearning); (2) Recency-Drop captures change of recency scores from Learning to SD (RecencyLearning – RecencySD); and (3) Primacy Progression captures how primacy accuracy progresses from Learning to LD (PrimacyLD -PrimacyLearning). We first entered both measures to explain risk of conversion to disease status from a) HC (N = 200) to MCI or AD; and b) from MCI (N = 353) to AD using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. We then contrasted the SPE-Index and SPE-Contrast scores to traditional total list-learning scores from the RAVLT as predictors of conversion. In Aim 2 we performed whole-brain analyses and posited that performance of SPE-Index scores would be subserved by distinct brain regions relevant to learning and encoding. We hypothesized that relevant language and memory driven SPE scores would be associated with corresponding neuroanatomical correlates. For example, we predicted that SPE-Primacy scores would be positively correlated with hippocampal, medial temporal, and frontal lobe regions given their associations with semantic encoding and retrieval. Alternately, we hypothesized that SPE-Recency scores would be positively correlated with inferior parietal and superior temporal gyrus regions, which would explain preserved phonological processing of items. Findings supported our Aim 1 hypothesis. We demonstrated that in preclinical (HC) individuals, diminished Short Delay recall performance of a word-list task best explained conversion risk. The Primacy item recall at Short Delay emerged as a particularly sensitive predictor of progression along the clinical AD spectrum for preclinical individuals. Total-RAVLT list scores at Short Delay also emerged as a predictor, although subsequent analyses highlighted that the primacy items at Short Delay were driving this signal. To our knowledge, these data are the first to support the importance of short delay primacy items as a predictor of disease development in the preclinical population. In MCI, all SPE-Contrast profiles in addition to all SPE-Index scores, with the exception of Recency at SD, significantly explained risk of progression to AD. In MCI, the SPE scores’ utility was similar to that of Total-RAVLT list scores. Findings for Aim 2 were mixed. Contrary to our hypotheses, in HC, we found Recency at Long Delay to be associated with left medial orbitofrontal cortical thickness, but no other significant SPE-Index or Total-RAVLT list score to have significant cortical volume or thickness correlates. By the MCI stage, the SPE-Index measures of Primacy and Middle list positions were associated with a range of regional volumes and thicknesses as were Total-RAVLT scores. The SPE-Index scores at this stage of disease did associate with more specific regions than Total-Scores; but SD, which was of primary interest in preclinical individuals, did not emerge with any significant SPE-Index correlates. Together, this study demonstrated that well-selected, theoretically driven neuropsychological measures can play a prominent role in identifying healthy individuals at great risk of developing AD. Importantly, the initial primacy items of long word-lists rely on semantic processing to be encoded. We propose that future study of other biomarkers to associate with SPE-Primacy in healthy individuals will be critical in order to capitalize on its sensitivity as a predictor of future disease. Furthermore, these SPE scores have the benefit of drawing on theoretical underpinnings and mapping on to specific AD disease processes that may be missed by total scores