Chen Y, Huang S, Chen T, Liang D, Yang J, Zeng C, Li X, Xie G, Liu Z: Machine Learning for Prediction and Risk Stratification of Lupus Nephritis Renal Flare. American Journal of Nephrology DOI 10.1159/000513566
Flare-ups (relapses) of lupus nephritis (LN) are common and when frequent represent a major factor contributing to a poor long-term outcome. Predicting and stratifying risk for such renal flares are very important factors that must be taken into consideration for optimal management of LN. Hopefully, such abilities will lead to better treatment and improved outcomes.
Chen and co-workers utilized machine learning protocols to develop a risk-score prediction model by analyzing derivation (n = 1,186) and internal validation (n = 508) cohorts of Chinese patients with biopsy-proven LN who had achieved a remission (complete or partial) following immunosuppressive therapy. Lupus renal flares were common in both cohorts (about 39%).
A 6-variable renal flare prediction model was developed and validated. These variables included: partial remission, endocapillary hypercellularity at baseline, age, serum albumin level, anti-dsDNA antibody level, and serum C3 at the time of remission. The C-statistic for the model prediction was about 0.82 (95% CI = 0.77–0.86).
While this study still needs independent external validation and may be affected by ancestry, the nature of therapy used to induce and maintain remission, and other factors, it remains a “proof-of-principle” for the value of machine learning systems in providing a more quantitative and personalized approach to renal flare risk prediction and stratification. It might lead to better outcomes if externally validated and applied to patients with LN who obtain a remission following immunosuppressive therapy.