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.

关于临床病理学特征在狼疮性肾炎 (LN) 预后的早期预测中的作用仍存在不确定性。确定完成常规随访的系统性红斑狼疮 (SLE) 患者，并对符合条件的病例进行回顾性审查。不良预后定义为全因死亡率或 eGFR 持续下降超过基线水平的一半或进展为终末期肾病 (ESRD)。构建了一个最佳的 Cox 回归模型，用于早期预测 LN 的不良预后。在 2163 例 SLE 患者中，376 例符合条件的 LN 病例入组，中位随访时间为 55 [27.0, 87.0] 个月。男女比例为 1:7.2，37 名患者 (9.8%) 进展至复合终点。ISN/RPS 类别与蛋白尿水平显着相关（磷- 值 < 0.001)，IV/IV + V 级患者，但不是 V 级患者，有最严重的蛋白尿。我们的最佳多变量 Cox 回归模型表明，性别、ISN/RPS 分级、肾小管萎缩/间质纤维化、血清白蛋白、蛋白尿三分位数及其相互作用与预后不良独立相关。ROC 分析和外部验证表明，我们的模型对于区分预后不良的 LN 患者是有效且稳健的。我们的研究构建了一个稳健的早期预测模型，以便在临床实践中识别 LN 患者的不良预后。我们发现蛋白尿和血清白蛋白在预测不良预后方面存在显着的交互作用。