A novel longitudinal approach in individual subjects to investigate the movement of tau over time using graph theory in clinical and preclinical stages of Alzheimer’s disease

Background We previously introduced an index characterizing inter‐regional tau deposition variability based on graph theory. Extending our method to additionally characterize the longitudinal tau spread for individual subjects based on directed graph theory, we examined the difference between longitudinal network measures and the regional tau deposit measures over time with respect to 1) differentiation of patients with AD, MCI, and cognitively unimpaired(CU) subjects, 2) correlation with memory measures in Aβ+ CU, and 3) sample size needed for a prevention trial for Aβ+/‐ CU. Method The longitudinal tau PET data is from 13 AD, 42 MCI, and 72 CU in ADNI . Using cerebellar grey reference region, we calculated the difference over two time points (1.1 years ± 0.2 (0.6‐1.6)) for the entorhinal, inferior temporal and metaROI SUVR, and the difference for indices, in‐strength and out‐strength each based on the directed tau longitudinal network using predefined nodes(inferior temporal and limbic). We examined regional/network based indices in term of the group differences, their correlation with AVLT‐LTM change. In addition, we also calculated sample size for a prevention clinical trial for Aβ+/‐ subjects separately with 80% power and 25% treatment effect. Result The longitudinal tau network measures(instrength and outstrength) are significantly different between the three groups (p<0.004) but not the SUVR change at entorhinal(p=0.62), inferior temporal(p=0.36) or tau meta ROI rates(p=0.21). In CU Aβ+ subjects, limbic outstrength significantly correlated to AVLT‐LTM rate(rs=‐0.52, p=0.013). The most significant regional value(tau metaROI) rate is not significantly correlated with AVLT LTM rate(rs=‐0.25,p=0.26). For a clinical trial, the Inferior temporal in‐out strength(net influx) requires a sample size of 584 subjects in CU Aβ+ group and a sample size of 26448 subjects in CU Aβ‐ Entorhinal tau SUVR, inferior temporal and tau metaROI SUVR needs a sample size for CU Aβ+ is 1623, 2450,2629 respectively. For CU Aβ‐ subjects, the sample size is 53537 for entorhinal SUVR, 50605 for inferior temporal SUVR and 916184 for tau metaROI SUVR Conclusion The network based longitudinal indices provide better power for differentiating patients at different stages, for examining tau’s effects on memory and for detecting a treatment effects in prevention trial.