Jie Liu, Xiang Zhou, Huijuan Zhu, Lian Duan, Linjie Wang, Yuxing zhao, Qi Zhang, Ling Qiu, Y. Zhang, Kan Deng, Di Zhang, Jing Liu, Jifang Liu, Wan Su, Zhang Ye, Yong Yao, Lin Lu
OBJECTIVE: ACTH-dependent Cushing's syndrome (CS) causes profound immune dysfunction and severe infections. This study aimed to characterize immune cell phenotypes, develop predictive models for differentiating Cushing's disease (CD) from ectopic ACTH syndrome (EAS) and predict infection risk, and evaluate immune recovery after surgical remission. DESIGN: Retrospective single-center study. METHODS: We studied 211 patients with ACTH-dependent CS (173 CD, 38 EAS) aged 11-75 years. Twelve patients were evaluated longitudinally after remission. Lymphocyte subsets were analysed by flow cytometry, severe infection incidence was documented, and machine learning models were developed for disease differentiation. RESULTS: EAS patients showed more severe immune suppression than CD, with reduced CD4+ T cells (224 [158.0-390.0] vs. 442 [313.0-611.0] cells/μL, P < .001) and CD3+ T cells (505 [342.0-815.0] vs. 930 [713.0-1246.0] cells/μL, P < .001), while elevated percentage of CD19+ B cells (18.2% [13.3%-25.5%] vs. 9.6% [6.1%-14.4%], P < .001). Machine learning-based model integrating immune and non-immune parameters achieved improved balanced accuracy of 75.4% (AUC = .840, 95% CI: 0.770-0.910) for CD/EAS differentiation. Severe infections occurred in 13.7% of patients. The multivariate model incorporating 24 h UFC, serum potassium nadir, CD19+ B cell percentage, and CD4+ T cell count predicted severe infection risk (AUC = .851, 95% CI: 0.774-0.927; sensitivity 86.2%, specificity 68.7%). Following surgical remission, profound T cell reconstitution occurred after 8 weeks, with CD4+ T cells increasing 3-fold (P < .001). CONCLUSIONS: Immune profiling distinguishes CD/EAS and predicts infection risk in ACTH-dependent CS. Surgical remission facilitates immune reconstitution.