Fan Shuang, Hong Haoyang, Lu Shengye, Wen Qi, Hong Shenda, Zhang Xiaohui, Xu Lanping, Wang Yu, Yan Chenhua, Chen Huan, Chen Yuhong, Han Wei, Wang Fengrong, Wang Jingzhi, Huang Xiaojun, Mo Xiaodong
Peking University People's Hospital, Peking University Institute of Hematology, National Clinical Research Center for Hematologic Disease, Beijing Key Laboratory of Hematopoietic Stem Cell Transplantation, Beijing, China.
National Institute of Health Data Science, Peking University, Beijing, China.
J Transl Int Med. 2025 Jun 20;13(3):253-266. doi: 10.1515/jtim-2025-0028. eCollection 2025 Jun.
Relapse is one of the most critical causes of transplant failure in patients with acute myeloid leukemia (AML) receiving haploidentical-related donor (HID) hematopoietic stem cell transplantation (HSCT). We aimed to develop an artificial intelligence (AI)-based predictive model for post-transplant relapse in patients with AML receiving HID HSCT.
This study included patients with consecutive AML (aged ≥ 12 years) receiving HID HSCT in complete remission (CR). We randomly selected 70% of the entire population ( = 665) as the training cohort for developing the model and nomogram, which were both evaluated using data from the remaining 30% of the patients (validation cohort, = 286). Furthermore, the model was validated in an independent cohort ( = 213) and in the clinical practice of five experienced clinicians.
Five variables (AML risk category, number of courses of induction chemotherapy for first CR, disease status, measurable residual disease before HSCT, and blood group disparity) were included in the final model (., PKU-AML model). The concordance index of the nomogram was 0.707. The Hosmer-Lemeshow test showed a good fit for this model ( = 0.205). The calibration curve was close to the ideal diagonal line, and decision curve analysis showed a significantly better net benefit for this model. The reliability of our prediction nomogram was demonstrated in a validation cohort, an independent cohort, and in clinical practice.
Our PKU-AML model can predict the relapse of patients with AML receiving HID HSCT in CR, providing an effective tool for the early prediction and timely management of post-transplant relapse.
复发是接受单倍体相合相关供者（HID）造血干细胞移植（HSCT）的急性髓系白血病（AML）患者移植失败的最关键原因之一。我们旨在开发一种基于人工智能（AI）的预测模型，用于预测接受HID HSCT的AML患者移植后的复发情况。
本研究纳入了连续的处于完全缓解（CR）期且接受HID HSCT的AML患者（年龄≥12岁）。我们随机选取全部患者的70%（n = 665）作为训练队列来开发模型和列线图，并使用其余30%患者的数据（验证队列，n = 286）对二者进行评估。此外，该模型在一个独立队列（n = 213）以及五位经验丰富的临床医生的临床实践中进行了验证。
最终模型（即PKU - AML模型）纳入了五个变量（AML风险类别、首次CR的诱导化疗疗程数、疾病状态、HSCT前的可测量残留病以及血型不相合情况）。列线图的一致性指数为0.707。Hosmer - Lemeshow检验显示该模型拟合良好（P = 0.205）。校准曲线接近理想对角线，决策曲线分析表明该模型的净效益显著更好。我们的预测列线图在验证队列、独立队列以及临床实践中均显示出可靠性。
我们的PKU - AML模型能够预测处于CR期且接受HID HSCT的AML患者的复发情况，为移植后复发的早期预测和及时管理提供了一种有效工具。
