Role of clinicopathological features for the early prediction of prognosis in lupus nephritis

Ambiguities remain regarding the role of clinicopathological characteristics in the early prediction of the prognosis of lupus nephritis (LN). Systemic lupus erythematosus (SLE) patients who completed routine follow-up were identified and retrospectively reviewed for eligible cases. Poor prognosis was defined as all-cause mortality or a persistent decrease of eGFR greater than half the baseline level or progression to end-stage renal disease (ESRD). An optimal Cox regression model was constructed for the early prediction of a poor prognosis for LN. Among the 2163 SLE patients, 376 eligible LN cases were enrolled in the study, with a median follow-up time of 55 [27.0, 87.0] months. The male-to-female ratio was 1:7.2, and 37 patients (9.8%) progressed to the composite endpoint. The ISN/RPS class was significantly associated with proteinuria levels (P-value < 0.001), and class IV/IV + V patients, but not class V patients, had the most severe proteinuria. Our optimal multivariate Cox regression model indicated that sex, ISN/RPS class, tubular atrophy/interstitial fibrosis, serum albumin, tertiles of proteinuria, and their interaction were independently associated with a poor prognosis. ROC analysis and external validation demonstrated that our model was efficient and robust for distinguishing LN patients with a poor prognosis. Our study constructed a robust and early predictive model for convenience in clinical practice to identify poor prognosis in LN patients. We found a significant interaction effect between proteinuria and serum albumin for the prediction of poor prognosis. LN patients with low-level proteinuria and hypoalbuminemia exhibit an increased hazard of progression to poor outcomes.