Spondyloarthritis & psoriatic arthritis
Predicting disease flares in axial spondyloarthritis using different machine learning models in the METEOR-SpA registry
Authors' conclusion
A reduced logistic-regression model showed moderate internal and modest external discrimination for short-term flare risk in axSpA.
Study context
- Population
- Patients with axial spondyloarthritis (axSpA) receiving bDMARDs in the METEOR-SpA registry
- Participants
- 1,167 patients3,865 visits
- Comparison
- Internal cross-validation (CV) vs GLAS
- Outcome
- Flare: ASDAS increase ≥0.9 (or Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) increase ≥2) at the next 6-month assessment
- Follow-up
- Next 6-month assessment
Methodological detailsmachine learning, patient-grouped cross-validation, decision curve analysis
- Machine learning
- Patient-grouped cross-validation
- Decision curve analysis
Performance
As reported in the abstract
Discrimination of the final ridge logistic regression model for flare at next 6-month assessment
| Test or model | ROC-AUC (95% CI) | Average precision (AP) |
|---|---|---|
| Internal CV | 0.72 (0.70 to 0.75) | 0.291 (0.245 to 0.337) |
| GLAS | 0.63 (0.59 to 0.68) | 0.142 (0.116 to 0.171) |
ROC-AUC (95% CI)
Internal CV0.72 (0.70 to 0.75)
GLAS0.63 (0.59 to 0.68)
Average precision (AP)
Internal CV0.291 (0.245 to 0.337)
GLAS0.142 (0.116 to 0.171)
Further findings
Patients with ≥1 flare, among 618 (52.9%) with ≥1 visit of known flare status
- Experienced ≥1 flare
- 170 (27.5%)
Predictors retained in the final model
- Eight predictors associated with flare: Axial Spondyloarthritis Disease Activity Score (ASDAS) (current and previous), female sex, flare history, enthesitis, infliximab exposure, PGA and CRP
- 8
Decision curve analysis
- Modest positive net clinical benefit at low threshold probabilities
Original publication
Predicting disease flares in axial spondyloarthritis using different machine learning models in the METEOR-SpA registry. RMD Open. Published 24 September 2026. doi:10. 1136/ rmdopen-2026-007147
Composed by AI from the published abstract and checked against it. How Briefs are made →