Abstract
Although network neuropsychology is a promising approach to the study of clinical profiles, the link between Alzheimer's disease (AD) biomarkers and neuropsychological networks is still undetermined. We hypothesized that network differences would exist between biomarker-positive and biomarker-negative participants, and that these would be driven by network nodes corresponding to performance on tests of episodic memory, as this is the cognitive domain most distinctively affected by AD since the earliest clinical stages. In this case-control study, we investigated sub-cohorts of individuals who had been (i) enrolled in the National Alzheimer's Coordinating Center initiative and (ii) tested with Version 3 of the Uniform Data Set neuropsychological battery (i.e. consisting of 11 tests). These included 1263 'β-amyloid positive' (A+), 1594 'β-amyloid negative' (A-), 442 'β-amyloid and hyperphosphorylated tau positive' (A + T+) and 734 'β-amyloid and hyperphosphorylated tau negative' (A-T-) participants. We first calculated neuropsychological residuals by regressing out age, years of education, sex, Clinical Dementia Rating scores and timepoint distance between neuropsychological and biomarker assessment. Secondly, we used rank-based correlations to define conditional associations across all pairs of test scores (i.e. the nodes of the network). Thirdly, we imposed a penalty (i.e. via the Least Absolute Shrinkage and Selection Operator method) to control for network sparsity. We then tested for differences in global network metrics and node centrality between A+ and A- and between A+T+ and A-T- participants using permutation-based inferential models. Differences were found between biomarker-positive and biomarker-negative sub-cohorts in global network metrics but, contrarily to our hypothesis, no differences were found in relation to episodic memory nodes. A significant node difference, however, was instead found in relation to category fluency (i.e. a test of semantic memory), with increased centrality observed among A+ participants. A similar, yet non-significant trend was also observed between A+T+ and A-T- participants. Network neuropsychology can complement and expand the study of cognitive performance carried out via 'traditional' univariate approaches. While univariate analyses reveal episodic memory decline in people with AD, this is not accompanied by any abnormalities at a neuropsychological network level. Our findings, however, highlight the importance of semantic memory alterations in A+ individuals. The wide set of neural and cognitive resources that sustain semantic memory may play a supportive role in the presence of neuropathology.