Longitudinal biomarker studies in human neuroimaging: capturing biological change of Alzheimer's pathology

Abstract

Despite extensive research, open questions about the biological underpinnings of Alzheimer’s disease (AD) remain. Neuroimaging biomarkers based on positron emission tomography (PET) and magnetic resonance imaging (MRI) offer in vivo insights into these complex biological changes and interactions. However, most evidence to date comes from cross-sectional studies, limiting our understanding of disease progression. Longitudinal studies enable the investigation of biological changes within individuals, revealing how pathology evolves over time. With this review, we provide an overview of how longitudinal imaging biomarker studies have advanced the field and how they can contribute to future research. We highlight longitudinal biomarker studies that have provided critical insights into disease trajectories, staging, and individual variability. We further assess longitudinal multimodal studies which have elucidated interactions between AD-specific pathology, amyloid-β and tau, and broader biological changes like neurodegeneration, neuronal dysfunction, vascular disease, and inflammation. Further, we discuss associations of brain changes with symptomatology and clinical outcomes and conclude with challenges and future directions.

Keywords: Alzheimer’s disease; Biomarker; Longitudinal; MRI; Neuroimaging; PET.

Fig. 1

Conceptual illustration of longitudinal biomarker dynamics in Alzheimer’s disease. A The influential model of Jack and colleagues [2] depicts archetypical sigmoidal curves representing isolated changes in Alzheimer’s disease biomarkers over time, based on the revised AT(N) framework. Adapted from [2]. B We propose that moving from isolated biomarker studies to longitudinal multimodal investigations can uncover more complex interactions and causal relationships between biomarkers. The curves shown in B are adapted from a longitudinal modeling study by Lattmann-Greve and colleagues [103], illustrating how multimodal longitudinal data can reveal intricate and interacting dynamics over time. In their study, the authors utilized longitudinal CSF, MRI, and cognitive scores in a multivariate probabilistic disease progression model to generate empirical biomarker disease progression curves. The resulting curves uncovered differential hypothetically implicated biomarker trajectories with cognition being preceded by morphometry and CSF-based Alzheimer’s disease biomarkers, respectively, and different timepoints of fastest change. The authors further assessed the relationship to change in fMRI encoding task activation. These changes in activation were nonlinear and independently associated with tau positivity and neurodegeneration. Adapted from [103]

Fig. 2

Proposed model of hyper- and hypoactivation in the Alzheimer’s disease pathological cascade. In Phase 0, non-pathological aging is characterized by functional changes (baseline, grey) in comparison with younger adults. Genetic predisposition to Alzheimer’s disease (AD) (i.e. APOE4 genotype) may cause a prolonged state of increased activation across mid- to late life (red dotted line). In Phase I, age- and/or genetic-related functional changes predispose certain regions to pathology accumulation (i.e. hyperphosphorylated tau in medial temporal lobe (MTL) and Aβ in posteromedial cortex (PMC)). This pathology accumulation coincides with the emergence of task-based hyperactivation (red), defined as increased activation contrasted against healthy older adults, which is evident when probed with episodic memory tasks. Hyperactivation first occurs in the hippocampus, particularly within dentate gyrus/CA3, due to tau-related perforant path degeneration (see inset box) and in PMC regions due to Aβ-related effects. Overt memory impairment is not yet evident at this stage. In Phase II, disconnection between the MTL and PMC results in exaggerated hyperactivation, as well as accelerated expansion of pathology in a vicious cycle. This peak of hyperactivation is associated with SCD and early MCI. In Phase III, a tipping point of high levels of tau pathology ultimately leads to neuronal silencing and neurodegeneration, resulting in hypoactivation (blue) which first emerges in the hippocampus and PMC. Simultaneously, a shift in hyperactivation to other regions (e.g. frontal cortex) occurs. Finally, in Phase IV, widespread pathology and neurodegeneration leads to further hypoactivation that encompasses large-scale cortical regions and networks, resulting in overt cognitive impairment characteristic of AD dementia. Adapted from [17]

Fig. 3

Conceptual illustration of the insights gained from longitudinal multimodal imaging biomarker studies on temporal, spatial, and causal aspects of Alzheimer’s disease pathology. This figure is not meant to be exhaustive but serves to illustrate the complex interplay of Alzheimer’s disease imaging biomarkers over time. A Temporal trajectories and relationships of imaging biomarkers across the disease continuum, derived from longitudinal studies discussed in this review paper. Curves depict the estimated onset, rate of change, and plateau phases for biomarkers. Dotted lines indicate biomarkers where limited longitudinal data is available. B Arrows depict shifts in biomarker trajectories influenced by inflammation and vascular disease, emphasizing how these additional factors alter disease trajectory. C Spatial correspondence of pathological processes across brain regions, illustrating patterns of co-localization and divergence as assessed by La Joie and colleagues [55] using multimodal longitudinal imaging biomarkers, can provide valuable insight into disease dynamics. Adapted from [55]. Brain plots from [55]. Reprinted with permission from AAAS. D Graph of causal relationships between imaging biomarkers based on studies reviewed above. Nodes represent distinct pathological processes implicated in Alzheimer’s disease. Directed edges indicate putative causal influences between processes, as estimated from longitudinal observational and experimental data to date as discussed in this review