Disease Phenotypes and Prediction of Outcome in ANCA-associated vasculitis
Författare
Summary, in English
Objectives
Anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) is a rare systemic autoimmune disease characterised by inflammation and destruction of small blood vessels. With research into the disease suffering from small sample sizes, this thesis aims to (1) address research data fragmentation in AAV through the integration of real-world observational registries, (2) stratify patterns of symptoms at disease onset, and (3) build models for the prediction of disease outcome.
Methods
Data from six European vasculitis registries were integrated using two complementary approaches. First the registries were combined using Semantic Web technologies, allowing federated access to aggregated data through a dedicated web interface. The quality of the underlying data was explored, and the characteristics, treatments and disease outcome of European patients described. Secondly, data was pooled in a central data storage. Using the central data, model-based clustering was used to study and stratify the diverse phenotypic presentations at disease onset. Lastly, prognostic models for key disease outcomes were built using survival modelling.
Results
The federated integration was successful, although some data quality concerns were identified, allowing access to an unprecedented cohort size of 5282 patients. Symptomatology, type of treatments used, mortality rates and rates of end-stage kidney disease were highly variable between the participating registries. Using model-based clustering, five clusters were identified, with distinct phenotypes, biochemical presentations, and disease outcomes – primarily stratified by kidney impairment and systemic inflammation. Building predictive models for disease outcome, known predictors of disease outcome were reidentified and compiled into comprehensive models, outperforming existing models in terms of predictive accuracy.
Conclusion
This thesis presents the first successful federated integration of distributed vasculitis datasets, allowing access to a cohort of unprecedented size. It further reinforces that AAV is beyond a binary construct and that the disease heterogeneity may be better described by five subcategories. While accurate prediction of disease outcome at the time of diagnosis is possible, the benefit of implementation of prediction models for the guidance of clinical decision-making needs further evaluation.
Anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV) is a rare systemic autoimmune disease characterised by inflammation and destruction of small blood vessels. With research into the disease suffering from small sample sizes, this thesis aims to (1) address research data fragmentation in AAV through the integration of real-world observational registries, (2) stratify patterns of symptoms at disease onset, and (3) build models for the prediction of disease outcome.
Methods
Data from six European vasculitis registries were integrated using two complementary approaches. First the registries were combined using Semantic Web technologies, allowing federated access to aggregated data through a dedicated web interface. The quality of the underlying data was explored, and the characteristics, treatments and disease outcome of European patients described. Secondly, data was pooled in a central data storage. Using the central data, model-based clustering was used to study and stratify the diverse phenotypic presentations at disease onset. Lastly, prognostic models for key disease outcomes were built using survival modelling.
Results
The federated integration was successful, although some data quality concerns were identified, allowing access to an unprecedented cohort size of 5282 patients. Symptomatology, type of treatments used, mortality rates and rates of end-stage kidney disease were highly variable between the participating registries. Using model-based clustering, five clusters were identified, with distinct phenotypes, biochemical presentations, and disease outcomes – primarily stratified by kidney impairment and systemic inflammation. Building predictive models for disease outcome, known predictors of disease outcome were reidentified and compiled into comprehensive models, outperforming existing models in terms of predictive accuracy.
Conclusion
This thesis presents the first successful federated integration of distributed vasculitis datasets, allowing access to a cohort of unprecedented size. It further reinforces that AAV is beyond a binary construct and that the disease heterogeneity may be better described by five subcategories. While accurate prediction of disease outcome at the time of diagnosis is possible, the benefit of implementation of prediction models for the guidance of clinical decision-making needs further evaluation.
Publiceringsår
2024
Språk
Engelska
Publikation/Tidskrift/Serie
Lund University, Faculty of Medicine Doctoral Dissertation Series
Avvikelse
2024:134
Fulltext
Dokumenttyp
Doktorsavhandling
Förlag
Lund University, Faculty of Medicine
Ämne
- Clinical Medicine
Nyckelord
- ANCA-associated vasculitis
- Epidemiology
- Cluster
- Classification
- Prediction
- Outcome
Aktiv
Published
Handledare
- Aladdin Mohammad
- Mårten Segelmark
- Mark A. Little
- Arthur White
ISBN/ISSN/Övrigt
- ISSN: 1652-8220
- ISBN: 978-91-8021-632-6
Försvarsdatum
8 november 2024
Försvarstid
09:00
Försvarsplats
Reumatologiska klinikens föreläsningssal, Lottasalen, Kioskgatan 5, Universitetssjukhuset i Lund. Join by Zoom: https://lu-se.zoom.us/j/63242456385
Opponent
- Raashid Luqmani (Professor)