Enhancing Cardiovascular Adverse Event Detection in ICI-Treated Cancer Patients: Lessons Learned from Natural Language Processing Integration with OMOP CDM
Cardiovascular (CV) adverse events (AEs) in cancer patients receiving immune checkpoint inhibitors (ICIs) are often under-detected in clinical trials, partly due to the strict inclusion criteria and limited follow-up typical of such studies. This underscores the need for real-world data analysis to gain a fuller understanding of these events.
Download our joint poster (as showcased at the OHDSI Global 2024 conference) by completing the form.
Authors: Clara L. Oeste1, Danielle Delombaerde2, Annelies Verbiest3, Iege Bassez1, Philip Debruyne4, Christof Vulsteke2, Dries Hens1
Background
• Cardiovascular (CV) adverse events (AEs) in cancer patients receiving immune checkpoint inhibitors (ICIs) are often under-detected in clinical trials.
• Clinical trials typically have strict inclusion criteria and incomplete follow-up, highlighting the need for real-world data analysis.
• This study used both structured data and NLP-extracted unstructured EHR data.
• The integration of NLP with structured data enriched the OMOP CDM, presenting an important analytic use case for the OHDSI community.
Conclusion
• NLP enhances the detection of CV AEs in ICI-treated cancer patients, particularly for less common events like pericarditis (78%) and myocarditis (60%).
• The combination of NLP and structured data improves AE identification, with NLP contributing 32-78% of additional cases across various categories.
• High precision, recall, and F1 scores validate the accuracy of NLP, enabling more comprehensive follow-up and monitoring in oncology care.
Complete the form above to download the poster and discover the methods used and the results obtained.
1 LynxCare, Leuven, Belgium
2 Department of Oncology, General Hospital AZ Maria Middelares, Ghent, Belgium and Center for Oncological Research (CORE), University of Antwerp, Antwerp, Belgium
3 Department of Oncology Antwerp University Hospital, Antwerp, Belgium
4 Department of Medical Oncology, General Hospital AZ Groeninge, Kortrijk, Belgium
Background
• Cardiovascular (CV) adverse events (AEs) in cancer patients receiving immune checkpoint inhibitors (ICIs) are often under-detected in clinical trials.
• Clinical trials typically have strict inclusion criteria and incomplete follow-up, highlighting the need for real-world data analysis.
• This study used both structured data and NLP-extracted unstructured EHR data.
• The integration of NLP with structured data enriched the OMOP CDM, presenting an important analytic use case for the OHDSI community.
Conclusion
• NLP enhances the detection of CV AEs in ICI-treated cancer patients, particularly for less common events like pericarditis (78%) and myocarditis (60%).
• The combination of NLP and structured data improves AE identification, with NLP contributing 32-78% of additional cases across various categories.
• High precision, recall, and F1 scores validate the accuracy of NLP, enabling more comprehensive follow-up and monitoring in oncology care.
Complete the form above to download the poster and discover the methods used and the results obtained.
1 LynxCare, Leuven, Belgium
2 Department of Oncology, General Hospital AZ Maria Middelares, Ghent, Belgium and Center for Oncological Research (CORE), University of Antwerp, Antwerp, Belgium
3 Department of Oncology Antwerp University Hospital, Antwerp, Belgium
4 Department of Medical Oncology, General Hospital AZ Groeninge, Kortrijk, Belgium
