Characterizing heterogeneity in the progression of Alzheimer's disease using longitudinal clinical and neuroimaging biomarkers

Abstract

Introduction Models characterizing intermediate disease stages of Alzheimer's disease (AD) are needed to inform clinical care and prognosis. Current models, however, use only a small subset of available biomarkers, capturing only coarse changes along the complete spectrum of disease progression. We propose the use of machine learning techniques and clinical, biochemical, and neuroimaging biomarkers to characterize progression to AD. Methods We used a large multimodal longitudinal data set of biomarkers and demographic and genotype information from 1624 participants from the Alzheimer's Disease Neuroimaging Initiative. Using hidden Markov models, we characterized intermediate disease stages. We validated inferred disease trajectories by comparing time to first clinical AD diagnosis. We trained an L2-regularized logistic regression model to predict disease trajectory and evaluated its discriminative performance on a test set. Results We identified 12 distinct disease states. Progression to AD occurred most often through one of two possible paths through these states. Paths differed in terms of rate of disease progression (by 5.44 years on average), amyloid and total-tau (t-tau) burden (by 10% and 69%, respectively), and hippocampal neurodegeneration ( P < .001). On the test set, the predictive model achieved an area under the receiver operating characteristic curve of 0.85. Discussion Progression to AD, in terms of biomarker trajectories, can be predicted based on participant-specific factors. Such disease staging tools could help in targeting high-risk patients for therapeutic intervention trials. As longitudinal data with richer features are collected, such models will help increase our understanding of the factors that drive the different trajectories of AD.

Predicting Alzheimer's disease progression using deep recurrent neural networks

“…However, the training procedure requires participants to have two timepoints, thus “wasting” data from participants with less or more than two timepoints. Therefore, state-based models (e.g., discrete or continuous state Markov model) that do not constrain the shapes of the biomarker trajectories or assume a fixed number of timepoints might be more suitable for this longitudinal prediction problem ( Sukkar et al, 2012 ; Goyal et al, 2018 ). Here, we considered recurrent neural networks (RNNs), which allow an individual’s latent state to be represented by a vector of numbers, thus providing a richer encoding of an individual’s “disease state” beyond a single integer (as in the case of discrete state hidden Markov models).…”

“…First, the “preprocessing” approach handles the missing data issue in a separate preprocessing step, by imputing the missing data (e.g., using the missing variable’s mean or more sophisticated machine learning strategies; Azur et al, 2011 ; Rehfeld et al, 2011 ; Stekhoven and Buhlmann, 2011 ; White et al, 2011 ; Zhou et al, 2013 ), and then using the imputed data for subsequent modeling. Second, the “integrative” approach is to integrate the missing data issue directly into the models or training strategies, e.g., marginalizing the missing data in Bayesian approaches ( Marquand et al, 2014 ; Wang et al, 2014 ; Goyal et al, 2018 ; Aksman et al, 2019 ).…”

Harnessing repeated measurements of predictor variables for clinical risk prediction: a review of existing methods

“…However, dynamic prediction can cover a broader range of aims and methods than those of interest here. In particular, the methods covered in this review are distinct to those for addressing calibration drift [15] or modelling disease state transitions [16].…”

Disease Modelling of Cognitive Outcomes and Biomarkers in the European Prevention of Alzheimer’s Dementia Longitudinal Cohort

“…Various disease progression and sub-type approaches have been proposed and developed. These include survival and multi-state models for investigating transitions between disease states ( Hubbard and Zhou, 2011 ; Vos et al, 2013 ; van den Hout, 2016 ; Wei and Kryscio, 2016 ; Robitaille et al, 2018 ; Zhang et al, 2019 ); mixed effects models (linear, generalized, non-linear) that incorporate subject-specific random effects and can be extended to handle latent time shifts, random change points, latent factors, processes and classes, hidden states, and multiple outcomes ( Hall et al, 2000 ; Jedynak et al, 2012 ; Liu et al, 2013 ; Proust-Lima et al, 2013 ; Donohue et al, 2014 ; Samtani et al, 2014 ; Lai et al, 2016 ; Zhang et al, 2016 ; Geifman et al, 2018 ; Li et al, 2018 ; Wang et al, 2018 ; Lorenzi et al, 2019 ; Proust-Lima et al, 2019 ; Villeneuve et al, 2019 ; Younes et al, 2019 ; Bachman et al, 2020 ; Kulason et al, 2020 ; Raket, 2020 ; Segalas et al, 2020 ; Williams et al, 2020 ) and can be combined with models for event-history data ( Marioni et al, 2014 ; Blanche et al, 2015 ; Proust-Lima et al, 2016 ; Rouanet et al, 2016 ; Li et al, 2017 ; Iddi et al, 2019 ; Li and Luo, 2019 ; Wu et al, 2020 ); event-based models which attempt to model the pathological cascade of events occurring as the disease develops and progresses through disease stages ( Fonteijn et al, 2012 ; Young et al, 2014 ; Chen et al, 2016 ; Goyal et al, 2018 ; Oxtoby et al, 2018 ); and various clustering approaches for discovering risk stratification/disease progression groups and endotypes. For example, those based on hierarchical, partitioning and model-based clustering algorithms/methods ( Dong et al, 2016 ; Racine et al, 2016 ; Dong et al, 2017 ; ten Kate et al, 2018 ; Young et al, 2018 ).…”