Real-World Evidence

Use Case Description

There is great promise in the use of Real-World Data (RWD) to generate so-called Real-World-Evidence (RWE). RWD is pragmatically defined as any data that is gathered within the context of standard of care. RWD is considered of increasing importance because it enables researchers to go beyond data gathered throughout traditional randomized clinical trials (RCTs) or carefully designed prospective research projects. Within the context of this Use Case, RWE is defined as any evidence and/or insights that is generated using RWD.

The figure below introduces the vision of our use case (= ultimate long-term goal). We strive towards personalized ‘real-world’ insights related to disease progression by scaling-up the use of ‘high-dimensional RWD’. With ‘high-dimensional RWD", we refer in the context of this Use Case to the combination of tabular data (e.g. Electronic Health Records or features extracted from other data sources), time series (e.g. evoked potentials, electrocardiogram or time series resulting from remote monitoring with devices) and images (e.g. magnetic resonance imaging).

We are convinced that robust end-to-end pipelines can be delivered that can be leveraged across different clinical finalities (e.g. multiple sclerosis (MS) and cardiovascular disorders) –since the key challenges that are currently obstructing us from reaching personalized real-world insights are the same across many (if not all) clinical finalities. In this use case, we focus on overcoming the following 4 key challenges. 

1. RWD consists of different data types, is messy and heterogeneous

2. Scaling-up high-dimensional RWD is accompanied with severe privacy risks

3. Selecting the biomarkers of interest is complicated

4. Coping with high-dimensional trajectory data with different levels of missingness complicates 
   building trustworthy prediction models

We focus on developing 6 main AI ‘solutions’ (cfr ‘pieces of the puzzle’, see picture above) and integrate them into end-to-end pipelines that are robust to be leveraged across different clinical finalities. We focus on multiple key clinical disease areas such as MS and cardiovascular disorders.

Some numbers of interest:

• Worldwide, there are more than 3 million people suffering from MS

• Cardiovascular diseases are the leading cause of death globally. An estimated 17.9 million people died from cardiovascular disease in 2019, representing 32% of all global deaths.

Outputs & Resources

Publications

Simpson-Yap, S., de Brouwer, E., Kalincik, T., Rijke, N., Hillert, J., Walton, C., Edan, G., Moreau, Y., Spelman, T., Geys, L. and Parciak, T., 2020. First results of the COVID-19 in MS Global Data Sharing Initiative suggest anti-CD20 DMTs are associated with worse COVID-19 outcomes. Multiple Sclerosis Journal, 26(3_SUPPL), pp.48-49.

Peeters, L. M., Parciak, T., Walton, C., Geys, L., Moreau, Y., De Brouwer, E., ... & Rijke, N. (2020). COVID-19 in people with multiple sclerosis: a global data sharing initiative. Multiple Sclerosis Journal, 26(10), 1157-1162.

De Brouwer, E., Becker, T., Moreau, Y., Havrdova, E. K., Trojano, M., Eichau, S., ... & Peeters, L. (2020). Longitudinal modeling of MS patient trajectories improves predictions of disability progression. arXiv preprint arXiv:2011.04749.

Steve Simpson-Yap, Ashkan Pirmani, Tomas Kalincik, Edward De Brouwer, Lotte Geys, Tina Parciak, Anne Helme, Nick Rijke, Jan Hillert, Yves Moreau, Gilles Edan, Tim Spelman, Sifat Sharmin, Robert McBurney, Hollie Schmidt, Arnfin Bergmann, Stefan Braune, Alexander Stahmann, Amber Salter, Bruce Bebo, Anneke Van der Walt, Helmut Butzkueven, Serkan Ozakbas, Rana Karabudak, Raed Alroughani, Juan I Rojas, Ingrid Van Der Mei, Guilherme Sciascia do Olival, Melinda Magyari, Ricardo Alonso, Richard Nicholas, Anibal Chertcoff, Ana Zabalza, Georgina Arrambide, Nupur Nag, Annabel Descamps, Lars Costers, Ruth Dobson, Aleisha Miller, Paolo Rodriguez, Vesna Prchkovska, Giancarlo Comi, Liesbet Peeters - 2233 Updated results of the COVID-19 in MS global data sharing initiative: anti-CD20 DMTs deleterious for COVID-19 severity but interferons not protective among people with MS: BMJ Neurology Open 2024;4:.

Simpson-Yap, S., Pirmani, A., De Brouwer, E., Peeters, L. M., Geys, L., Parciak, T., ... & Kalincik, T. (2022). Severity of COVID19 infection among patients with multiple sclerosis treated with interferon-β. Multiple Sclerosis and Related Disorders, 66, 104072.

Simpson-Yap, S., De Brouwer, E., Kalincik, T., Rijke, N., Hillert, J. A., Walton, C., ... & Peeters, L. (2021). Associations of disease-modifying therapies with COVID-19 severity in multiple sclerosis. Neurology, 97(19), e1870-e1885.

Pirmani, A., De Brouwer, E., Geys, L., Parciak, T., Moreau, Y., & Peeters, L. M. (2023). The Journey of Data Within a Global Data Sharing Initiative: A Federated 3-Layer Data Analysis Pipeline to Scale Up Multiple Sclerosis Research. JMIR Medical Informatics, 11(1), e48030.

Khan, H., Geys, L., Baneke, P., Comi, G., & Peeters, L. M. (2024). Patient level dataset to study the effect of COVID-19 in people with Multiple Sclerosis. Scientific data, 11(1), 149.

De Brouwer, E., Becker, T., Werthen-Brabants, L., Dewulf, P., Iliadis, D., Dekeyser, C., ... & Peeters, L. (2024). Machine-learning-based prediction of disability progression in multiple sclerosis: An observational, international, multi-center study. PLOS digital health, 3(7), e0000533.

Pirmani, A., De Brouwer, E., Oldenhof, M., Passemiers, A., Van Wijmeersch, B., Butzkueven, H., Horakova, D., Ozakbas, S., Patti, F., Madueno, S.E. and Duquette, P., 2024. Transforming Multiple Sclerosis Research: Advancing Disability Progression Insights through Practical and Precise Federated Learning using Real-World Data.

Pirmani, A., Oldenhof, M., Peeters, L.M., De Brouwer, E. and Moreau, Y., 2024. Accessible ecosystem for clinical research (federated learning for everyone): development and usability study. JMIR Formative Research, 8, p.e55496.

Pirmani, A., Moreau, Y. and Peeters, L.M., 2024. Unlocking the Power of Real-World Data: A Framework for Sustainable Healthcare. Studies in Health Technology and Informatics, 316, pp.1582-1583.

Giraldo, D.L., Khan, H., Pineda, G., Liang, Z., Lozano-Castillo, A., Van Wijmeersch, B., Woodruff, H.C., Lambin, P., Romero, E., Peeters, L.M. and Sijbers, J., 2024. Perceptual super-resolution in multiple sclerosis MRI. Frontiers in Neuroscience, 18, p.1473132.

Parciak, M., Vandevoort, B., Neven, F., Peeters, L.M. and Vansummeren, S., 2024. Schema matching with large language models: an experimental study. arXiv preprint arXiv:2407.11852.

Pirmani, A., De Brouwer, E., Arany, Á., Oldenhof, M., Passemiers, A., Faes, A., Kalincik, T., Ozakbas, S., Gouider, R., Willekens, B. and Horakova, D., 2025. Personalized federated learning for predicting disability progression in multiple sclerosis using real-world routine clinical data. NPJ digital medicine, 8(1), p.478.

Faes A. Primani A., Moreau Y, Peeters, L.M., 2025. Applying Federated Learning to Block-Term Tensor Regression for Heart Disease Prediction. FLTA2025

PhD dissertations

Lorin Werthen-Brabants “Quantifying Uncertainty and Improving Reliability of Time-Series Based Deep Learning Models”, 2023

Edward de Brouwer “Machine learning modeling of time-dependent patient trajectories”, 2023

Ashkan Pirmani “Federated artificial intelligence for multiple sclerosis patient trajectories in a real-world setting”, 2024

Hamza Khan “New kids on the block: Improving the care of people with multiple sclerosis by introducing “Radiomics” and “Epomics””, 2025

Marcel Parciak “Accelerating Open Data Integration of Real-World Health Data Silos”, 2025

Collaborative projects

• Already existing:

  • MS DataConnect,

  • Covid19 in MS Global Data Sharing Initiative,

  • MS Data Alliance,

  • BELTRIMS

• New since kick-off new roadmap:

  • BRANDO,

  • FL4PHM,

  • FL4Cardio

Related projects and initiatives outside the Use Case

• OHDSI Belgium,

• Medemotion,

• Health Campus Limburg,

• ELIXIR BE connection,

• one million genome biobank,

• data4PHM

Multi-stakeholder partners

• Hospitals: e.g. Noorderhart, Ziekenhuis Oost-Limburg, …

• Industry: Janssen, Novartis, …

• Regulators: Health Data Agency, EWI, …

• …

Outreach

Used to showcase the successful collaborations between Belgium and Australia in the field of secondary re-use of health data during the Royal Economic Mission in Melbourne 
(18th October – 28th of October 2023)

Uptake / Proof of Concept

What's next

Phase 1: Extending FAIR's real-world evidence expertise to new clinical domains

Goal

Develop an end-to-end pipeline for cardiovascular diseases (POC1) and establish a sandbox for additional use cases (POC2), leveraging methodologies and experience from the Multiple Sclerosis use case developed during FAIR 2019-2023. The aim is to address similar challenges in cardiovascular diseases, new multiple sclerosis datasets, and other clinical application domains.

Methodology

Apply and adapt the methodologies developed within the FAIR 2019-2023 Multiple Sclerosis use case as a foundation for new clinical domains and datasets.

Results

Identification and documentation of first lessons learned from transferring FAIR's real-world evidence methodologies towards new clinical applications.

Phase 2: Developing and benchmarking new AI solutions

Goal

Develop new and improved AI solutions for real-world evidence generation, using FAIR Multiple Sclerosis tasks as a benchmark.

Methodology

Parallel development and benchmarking of the following AI solutions using FAIR Multiple Sclerosis tasks:

• Automated schema mapping for heterogeneous tabular healthcare data (POC3)

• Automated anomaly detection in real-world data (POC4)

• MRI harmonisation strategies (POC5)

• Privacy-preserving learning across multiple organisations (POC6)

• Biomarker identification and selection through visualisation of high-dimensional longitudinal datasets (POC7)

• Coping with IID versus non-IID bias in federated learning (POC8)

Results

New and improved AI solutions for:

• Tabular data handling through automated schema mapping and anomaly detection

• MRI harmonisation

• Privacy-preserving multi-party learning

• Biomarker identification and selection

• Federated learning under heterogeneous data distributions

In addition, an improved end-to-end pipeline will be developed for cardiovascular diseases (POC9) and other clinical use cases (POC10).

Phase 3: Evaluating and validating impact

Goal

Re-evaluate the impact of the revised end-to-end pipelines and newly developed AI solutions.

Methodology

Use the outcomes and performance of POC1 and POC2 as benchmarks for comparison against the revised solutions.

Results

• Validation of revised AI solutions and end-to-end pipelines

• Quantification of improvements and added value

• Identification and documentation of lessons learned for future real-world evidence applications

Team & Contact

Contact

Use Case Lead - Liesbet M. Peeters: e-mail

Scientific Coordinator - Axel Faes: e-mail

Use Case Lead(s)

Liesbet M. Peeters, Research Group of Biomedical Data Sciences, UHasselt

Principal Investigators / AI Expert Leads

Yves Moreau, KU Leuven; Stijn Van Summeren, UHasselt; Frank Neven, UHasselt; Jan Sijbers, imec / UAntwerpen; Celine Vens; Dirk Deschrijver, imec / UGent; Yvan Saeys, VIB

Postdoctoral Researchers

Axel Faes, UHasselt; Ashkan Pirmani, KU Leuven / UHasselt; Lorin Werthen-Brabants, imec / UGent

PhD Researchers

Marcel Parciak, UHasselt; Diana Girado; Jana Osstyn; Robbe Dhondt; Luca Leomazzi

Research Groups