Predicting abatacept retention using machine learning
Authors' conclusion
The gradient-boosting classifier model identified key patient features predictive of abatacept retention from this large, real-world study population. The SHAP values conveyed the directionality and importance of BMI, functional status, ACPA serostatus, Patient Global Assessment, and age for abatacept retention.
Study context
- Population
- Patients with rheumatoid arthritis (RA) initiating abatacept in the global, real-world ACTION and ASCORE trialstest set n = 1055
- Participants
- 5,320 patients
- Primary endpoint
- Prediction of 12-month abatacept treatment retention (treatment >365 days, or ≤365 days with remission or major clinical response by EULAR criteria)
- Follow-up
- 12 months
Methodological detailsmachine learning prediction study, 10 machine learning models, training/validation and test cohorts, SHAP values
- Machine learning prediction study
- 10 machine learning models
- Training/validation and test cohorts
- SHAP values
Primary result
As reported in the abstract
Prediction of 12-month abatacept retention, gradient-boosting classifier (best model)
Testing accuracy (training set n = 4218)
62%
AUC ROC in test set
0.620
95% CI 0.586, 0.653
F1 score in test set (0.625, 0.689)
0.659
Further findings
12-month abatacept retention and discontinuation
- Retention rate (n = 3236)
- 61%
- Discontinuation rate (n = 2037)
- 39%
Original publication
Predicting abatacept retention using machine learning. Arthritis Res Ther. Published 1 February 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →