肿瘤皮肤侵犯是头颈癌快速复发的独立预测因素
该研究回顾性分析246例接受手术及辅助治疗的头颈鳞状细胞癌(HNSCC)患者,将复发分为快速复发(RR,辅助治疗前)、短间隔复发(SIR,辅助治疗后≤6个月)和标准复发(SR,辅助治疗后>6个月)。多因素分析显示皮肤侵犯是RR的独立预测因素(OR=3.492,p=0.039),随机森林模型也确认皮肤侵犯等为预测因素。
要点
- 皮肤侵犯是快速复发的独立预测因素(OR=3.492)。
- 随机森林模型准确率93%、AUC 0.96。
- 快速复发指辅助治疗前即复发。
核心结论
- 皮肤侵犯是头颈鳞状细胞癌快速复发的独立预测因素。
研究设计与样本
- 回顾性10年研究,纳入246例接受手术及辅助治疗的HNSCC患者。
主要发现
- 246例中89例复发:SR 45例,SIR 27例,RR 17例。
- 多因素分析中皮肤侵犯是RR的独立预测因素(OR=3.492,p=0.039)。
- 随机森林模型显示皮肤侵犯、淋巴结状态、吸烟包年、肿瘤大小为预测因素(准确率93%,AUC 0.96,F1 0.93)。
临床怎么用
- 对伴有皮肤侵犯的HNSCC术后患者应警惕快速复发,需加强监测。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Head and neck squamous cell carcinoma (HNSCC) is an aggressive malignancy, with 50% of patients recurring. A subset of patients experience rapid recurrence (RR) postoperatively but prior to adjuvant therapy. This study identifies factors associated with RR and additional recurrence intervals: short-interval recurrence (SIR) and standard recurrence (SR). METHODS: Retrospective 10-year review of 246 HNSCC patients undergoing surgery with adjuvant therapy. Recurrence was categorized as RR (prior to initiation of adjuvant therapy), SIR (≤ 6 months post-adjuvant therapy), and SR (> 6 months post-adjuvant therapy). Univariate analysis (UVA), multivariate analysis (MVA), and machine learning Random Forest models were employed to identify predictors of each recurrence interval. RESULTS: Of the 246 patients, 89 recurred (45 SR, 27 SIR, 17 RR). On MVA, skin invasion (OR = 3.492, p = 0.039) was a unique predictor of RR. Random Forest feature importance also revealed skin invasion, along with nodal status, tobacco pack-years, and tumor size as predictors with strong performance (accuracy 93%, AUC 0.96, F1 0.93). CONCLUSION: Skin invasion is a unique independent predictor of RR, confirmed by two statistical models. These patients warrant further study.
引用
- Memon AA, Khalil M, Espinosa OV, et al. Tumoral Skin Invasion Is an Independent Predictor of Rapid Recurrence in Head and Neck Cancer. Head Neck. 2025. PMID: 41054923. doi:10.1002/hed.70066
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。