关键词: 系统性红斑狼疮, 自身抗体, 器官受累, 风险预测模型, logistic 回归

Abstract:

Objective To develop organ-specific risk prediction models for organ involvement in systemic lupus erythematosus （SLE） based on multidimensional autoantibody profiles， and to validate their performance using a time-split cohort， thereby providing a quantitative tool for early identification of high-risk patients. Methods A retrospective single-center cohort study was conducted with model development and validation performed in chronological order. The development cohort included 170 SLE patients， and 60 patients from the same institution were enrolled as a time-extrapolated validation cohort. Organ/system involvement was defined as a binary outcome according to predefined criteria， including serosal， renal， pulmonary， cutaneous， hematological， articular， and hepatic involvement. Candidate variables were screened by univariate analysis （χ2 test， t-test， or Mann?Whitney U test）， and independent associated factors were identified by multivariate logistic regression. Model discrimination was assessed using ROC analysis， AUC， sensitivity， specificity， and Youden index； robustness was evaluated by stratified 10-fold cross-validation. Calibration was assessed with the Hosmer-Lemeshow test and Brier score. Clinical utility was evaluated by decision curve analysis （DCA）. Results Among 170 SLE patients， the prevalence of organ involvement was： hematological 79.41%， renal 51.76%， pulmonary 51.17%， cutaneous 38.82%， articular 33.53%， hepatic 19.18%， and serosal 18.02%. Multivariate regression demonstrated that anti-P0 and anti-NUC were associated with serosal involvement； anti-Ro52 was associated with renal involvement； anti-dsDNA， ln（age）， and ln（disease duration） were associated with cutaneous involvement； anti-nRNP/Sm was associated with hematological involvement. Anti-NUC was an independent risk factor for articular involvement， whereas anti-AMA-M2 was an independent protective factor. The AUC of individual organ-specific models ranged from 0.60 to 0.80. The overall model achieved an AUC of 0.88 （95%CI： 0.81 - 0.94）， with a mean AUC of 0.86 （95%CI： 0.78 - 0.93） in stratified 10-fold cross-validation. Hosmer-Lemeshow tests in development and validation cohorts were > 0.10， overall Brier score ≈ 0.12. DCA showed a net clinical benefit of the models at threshold probabilities between approximately 10% and 50%. Conclusion Organ-specific risk prediction models based on multidimensional autoantibody profiles demonstrate favorable discrimination and acceptable calibration in a time-split validation cohort， indicating that autoantibody profiles can serve as a useful auxiliary tool for risk stratification of organ involvement in SLE patients.

Key words: systemic lupus erythematosus, autoantibody, organ involvement, risk prediction model, logistic regression