Liang Xiaojie, Lu Weixiang, Li Tong, Luo Baiwei, Wu Yuzhe, Lin Chaoran, Liu Yang, Wang Liang
Department of Hematology, Beijing Tongren Hospital, Capital Medical University, Beijing, 100730, China.
Stomatological Hospital, School of Stomatology, Southern Medical University, Guangzhou, 510515, China.
Ann Hematol. 2025 May 24. doi: 10.1007/s00277-025-06413-y.
Primary lymphoma of bone (PLB) is a rare extranodal lymphoma, and its epidemiology and prognosis remain controversial. We conducted a retrospective analysis of 1,222 patients with PLB in the Surveillance, Epidemiology, and End Results (SEER) database to investigate its epidemiology and prognostic factors. The incidence of PLB peaked in 1992 with an average annual percent change of 1.72 after a significant rise from 1975 to 1992, followed by a general decline. The risk of death from PLB involves both patient and treatment factors. Survival analysis revealed that age, stage, laterality, chemotherapy, and primary site significantly influence both overall survival and disease-specific survival. We integrated and compared 99 machine learning algorithms, and identified the Random Survival Forest (RSF) model as the most effective for predicting PLB outcomes. Patients were stratified into low- and high-risk groups according to the RSF model score. The incidence of PLB began to decrease after 1992, with variations by age, race, and gender. The factors influencing the prognosis of PLB are multifaceted. And the RSF model showed promising performance, aiding clinicians in early prognosis identification and improving clinical outcomes through revised management strategies and patient care.
原发性骨淋巴瘤（PLB）是一种罕见的结外淋巴瘤，其流行病学和预后仍存在争议。我们对监测、流行病学和最终结果（SEER）数据库中1222例PLB患者进行了回顾性分析，以研究其流行病学和预后因素。PLB的发病率在1992年达到峰值，从1975年到1992年显著上升后，年均变化率为1.72%，随后总体呈下降趋势。PLB的死亡风险涉及患者和治疗因素。生存分析显示，年龄、分期、病变侧别、化疗和原发部位对总生存和疾病特异性生存均有显著影响。我们整合并比较了99种机器学习算法，确定随机生存森林（RSF）模型是预测PLB预后最有效的模型。根据RSF模型评分将患者分为低风险和高风险组。1992年后PLB的发病率开始下降，存在年龄、种族和性别的差异。影响PLB预后的因素是多方面的。并且RSF模型表现出良好的性能，有助于临床医生早期识别预后，并通过修订管理策略和患者护理改善临床结局。
