The Predictive Power of Past Failure: Using Clinically Available Data to Improve Treatment Algorithms
MetadataShow full item record
CitationAnderson, Edwin. 2019. The Predictive Power of Past Failure: Using Clinically Available Data to Improve Treatment Algorithms. Master's thesis, Harvard Extension School.
AbstractPediatric rheumatology is a small but rapidly evolving specialty within the field of immunology that has been revolutionized by development of a new class of drugs called biologics. Treatment with these specialized agents results in clinical remission for 60-70% of juvenile idiopathic arthritis patients. The aim of this investigation is to determine if patient characteristics prior to an anti-TNF biologic’s initiation or data from the first three months of response can serve as predictors of the medication’s efficacy. This is a retrospective case-control study conducted on a local population of polyarticular JIA patients treated at Boston Children’s Hospital Rheumatology Program. I reviewed the records of 150 pediatric patients, each being treated with an anti-TNF agent as their first line biologic. Fifty patients discontinued treatment within their first year of prescription while the remaining 100 served as controls. As a secondary analysis, patients were stratified as either responders or nonresponders based on their documented reasons for switching off their first line anti-TNF, regardless of case/control status. Younger age at anti-TNF initiation, reduction in neutrophil count and reduction in JADAS10 score at 3 months showed significant predictive power in determining if a patient is likely to switch off their first line anti-TNF within the first year of treatment. Rheumatoid factor positivity and 3 month reduction in WBC showed significant predictive power in determining if an anti-TNF biologic was discontinued due to clinical ineffectiveness. This study demonstrates the value of mining existing patient medical records to guide future research and contribute to more personalized, effective treatment selection.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:42004088
- DCE Theses and Dissertations