Description
Date depot: 14 octobre 2022
Titre: Early prediction of neurodegenerative diseases using large transnational electronic health records databases for better prevention
Directeur de thèse:
Stanley DURRLEMAN (ICM)
Encadrant :
Thomas NEDELEC (ICM)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Sciences de l’information et sciences du vivant
Resumé: Context and motivations
The global incidence of neurodegenerative disease is quickly growing, mainly attributable to aging of the population. Dementia affected 10.5 million Europeans in 2015, and projections suggest that 13.5 million Europeans will be affected in 20301. This epidemic has a strong impact in terms of human suffering and monetary cost as it requires an important social care. Neurodegenerative diseases are progressive and take decades to become established2. This makes prevention key to design efficient treatments by treating the possible multiple causes of the disease. However, it remains difficult to identify these individuals. For example, testing for known biomarkers of early disease through (e.g.) performing PET scan or lumbar puncture to every individual over 65 is neither practical, ethical nor cost efficient.
Clinical health records may offer a direct advantage in terms of implementation and impact as it enables to run new complex longitudinal analysis and importantly to test multiple hypothesis while keeping some statistical significance. We recently discovered at the Aramis Lab (ICM, Paris) that some health conditions are associated with AD in a window of exposure from 2-10 years before AD’s diagnosis3.
Scientific Project
Our first objective is to see if GP diagnosed conditions can help to detect some patients at risk of dementia who would benefit from improved early primary and secondary prevention strategies. To this end, we aim to use transnational medical records (projected-based access to UK and France THIN databases, and permanent access to the French social security system database) to identify the patients who are the most at risk to develop neurodegenerative diseases, and to develop prediction algorithms that will support personalized early intervention and prevention measures. The thesis is funded by a JPND (EU Joint Programme – Neurodegenerative Research) grant corresponding to a consortium between the ICM, the Karolinska Institute and the University of Queensland. The goal of this consortium is to develop such prediction algorithms at the European level to build robust analyses.
Our second objective is to compare the different neurodegenerative diseases to better understand the specificity of each disease. We will undertake cross-disorder analyses in order to identify pre-clinical markers that are common to all neurodegenerative diseases, as well as those that are specific to each disorder. This will be leveraged for building disease-specific risk prediction. Our final objective is to analyze our prediction algorithms and use longitudinal data to understand why a given health condition is a good predictor, the ultimate goal being to separate between unspecific early symptoms and possible causes of the diseases.
Several machine learning algorithms (logistic regression, SVM, deep neural sequence transduction etc.) will be evaluated in terms of indicators of predictive performance (e.g., C-statistics, sensitivity, false positive, specificity). We will study whether the risk of developing a given neurodegenerative disease can be computed with sufficient precision. Knowing the performance of such algorithms for prevention trials and identification of patients at risk (precision, recall, etc.) will help to address the ethical and economic discussion for precision prevention of neurodegenerative diseases. It will also complement current approaches based on blood samples which also tries to predict risk of developing a specific neurodegenerative disease before the first symptoms.
1. Ansart M, Epelbaum S, Gagliardi G, et al. Reduction of recruitment costs in preclinical AD trials: validation of automatic pre-screening algorithm for brain amyloidosis. Stat Methods Med Res. 2020;29(1):151-164. doi:10.1177/0962280218823036
2. Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. The Lancet. 2020;396(10248):413-446. doi:10.1016/S0140- 6736(20)30367-6
3. Nedelec T, Couvy-Duchesne B, Monnet F, et al. Identifying health conditions associated with Alzheimer’s disease up to 15 years before diagnosis: an agnostic study of French and British health records. The Lancet Digital Health. 2022;4(3):e169-e178. doi:10.1016/S2589- 7500(21)00275-2
Doctorant.e: Guinebretiere Octave