SebastiaanRam / sem-mci-prediction

Abstract and figures for the poster presentation "Structural modeling of clinical factors for simultaneous validation and prediction of future conversion to mild cognitive impairment" at the Alzheimer's Initiative International Conference (AAIC) 2023

Structural modeling of clinical factors for simultaneous validation and prediction of future conversion to mild cognitive impairment

Sebastiaan R Ram¹, Bryan A Strange, PhD^2,3, Linda Zhang, PhD², Teodoro del Ser², Elizabeth Lucia Valeriano Lorenzo, MSc², Meritxell Valentí², María Ascensión Zea-Sevilla, MD. PhD⁴, Belén Frades², Tom Heskes¹, Pedro Larrañaga³, Concha Bielza³ and Pascual Sanchez-Juan, PhD, MD²

(1) Radboud University Nijmegen, Nijmegen, Netherlands, (2) CIEN Foundation, Queen Sofia Foundation Alzheimer Centre, Madrid, Spain, (3) Universidad Politecnica de Madrid, Madrid, Spain, (4) Instituto Neurológico Beremia, Madrid, Spain

Background:

Identifying risk factors of future conversion from cognitively normal to mild cognitive impairment is crucial for dementia prevention and understanding the clinical processes resulting in the development of cognitive deterioration. One major challenge faced with this task is that, when identifying these factors, not only their individual contribution must be determined but also their interplay with other indicators.

Methods:

To map the relationship between clinical risk factors and conversion to mild cognitive impairment, a set of structural equation and latent growth models was created and validated based on input from clinical experts in dementia research. These models are powerful as they allow the researcher to test hypotheses between observed variables and their underlying latent constructs. In addition to verification of the model structure, a method for predicting individual factors, such as mild cognitive impairment, was implemented on out-of-sample converters and controls directly on the model, a feature which is normally uncommon in structural equation modeling.

Results:

In a study of 1213 elderly participants we found that these structural models perform similar to current state-of-the-art logistic regression models in terms of Area-Under-the-Curve (AUC) score on a 1-year prediction (AUC $>$ 0.9), yet offer a significant increase in flexibility and insight into the direct and indirect effects of endogenous variables on conversion to mild cognitive impairment. We also observe an increase in predictive performance for both 4 and 8-year predictions in models fitted with an additional visit, compared to single-visit predictions on neuropsychological and MRI data alone (AUC increase +- 0.02). This allows us to monitor the development of cognitive health in subjects over time.

Conclusions:

The results from this study suggest a powerful technique for creating SEMs and LGMs using expert knowledge with the ability to perform robust path analysis, longitudinal monitoring of cognitive health and early prediction of future conversion to mild cognitive impairment. We expect that clinical experts can use this method to continue testing and validating new hypotheses in dementia research.