Machine Models Predict Renal Flares in Patients With LN
Research shows renal flare of lupus nephritis (LN) can be projected using an eXtreme Gradient Boosting (XGBoost) method model, as well as with a simplified risk score prediction model (SRSPM).
The SRSPM can also stratify flare risk, but both models are effective in individualized management and clinical decision making in patients with LN, the investigators found.
Researchers divided 1,694 patients with biopsy-proven LN who achieved remission after treatment into a derivation cohort of 1186 and an internal validation cohort of 508, at a ratio of 7:3.
Using an XGBoost method model—developed from 59 variables, including demographic, clinical, immunological, pathological, and therapeutic characteristics—and a derivative, SRSPM, developed from key variables chosen by XGBoost model, researchers analyzed the 5-year relapse rates and each model’s predictive performance. Both were found to have good predictive performance, totaling 39.5% in the derivation cohort and 38.2% in the internal validation cohort.
“The SRSPM comprised 6 variables, including partial remission and endocapillary hypercellularity at baseline, age, serum Alb, anti-dsDNA, and serum complement C3 at the point of remission. Using Kaplan-Meier analysis, the SRSPM identified significant risk stratification for renal flares,” the authors concluded.
--Angelique Platas
Reference:
Chen Y, Huang S, Chen T. et al. Machine learning for prediction and risk stratification of lupus nephritis renal flare. Am J Nephrol. Published 2021