头颈癌放射性甲状腺功能减退累积风险模型与评估工具的多中心开发与验证
该多中心回顾性研究纳入1407例头颈癌放疗患者,开发并外部验证了一种整合CT放射组学、剂量组学、剂量体积直方图参数及临床变量的多模态Cox模型,用于个体化累积放射性甲状腺功能减退(RIHT)风险评估。模型在开发队列和三个外部验证队列中均表现出一定的区分度,并据此划分出四个时间风险组,组间无RIHT生存率差异显著。
要点
- 多模态Cox模型整合放射组学、剂量组学、平均甲状腺剂量、年龄和N分期。
- 模型在开发队列C-index/总体AUC为0.714/0.765,外部验证队列为0.725/0.800、0.680/0.754、0.627/0.690。
- 四个时间风险组的RIHT事件率分别为21.7%、46.1%、58.3%、71.8%,组间无RIHT生存差异显著(log-rank p < 0.001)。
核心结论
- 外部验证的多模态Cox模型可利用常规放疗计划数据提供时间分辨的个体化RIHT风险估计和风险分层。
研究设计与样本
- 回顾性多中心研究,纳入四个队列共1407例患者,在训练队列开发模型并在三个独立队列进行外部验证。
主要发现
- 1407例患者平均年龄55.4±13.8岁,男性1027例,RIHT发生694例(49.3%)。
- 最终模型整合三个CT放射组学特征、四个剂量组学特征、平均甲状腺剂量、年龄和N分期。
- 模型在开发队列C-index/总体AUC为0.714/0.765,三个外部验证队列分别为0.725/0.800、0.680/0.754和0.627/0.690。
- 四个时间风险组的RIHT事件率分别为21.7%、46.1%、58.3%和71.8%,无RIHT生存差异显著(log-rank p < 0.001)。
临床怎么用
- 模型已封装为开源软件工具,可用于基于常规放疗计划数据的个体化RIHT风险评估。
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
OBJECTIVES: Radiation-induced hypothyroidism (RIHT) is a common late toxicity after head and neck radiotherapy. We aimed to develop and externally validate a multimodal Cox model for patient-specific cumulative RIHT risk assessment. MATERIALS & METHODS: This retrospective multicentre study included 1407 patients across four cohorts. The model was developed in the training cohort and externally validated in three independent cohorts. Candidate predictors included CT radiomics, dosiomics, dose-volume histogram parameters, and clinical variables. A multimodal Cox model was developed and used to estimate cumulative risks across time horizons. Temporal risk groups were derived from predicted 3- and 7-year risks. RESULTS: A total of 1407 patients (mean age, 55.4 ± 13.8 years; 1027 men) were evaluated and RIHT occurred in 694 (49.3%). The final model integrated three CT radiomics features, four dosiomics features, mean thyroid dose, age and N stage. It achieved C-index/overall AUC values of 0.714/0.765 in the development cohort and 0.725/0.800, 0.680/0.754, and 0.627/0.690 in the three external validation cohorts. Four temporal risk groups had RIHT event rates of 21.7%, 46.1%, 58.3%, and 71.8%, with significant differences in RIHT-free survival (log-rank p < 0.001). The model was packaged as an open-source software tool for individualised risk assessment. CONCLUSION: The externally validated multimodal Cox model provided time-resolved, patient-specific RIHT risk estimates and temporal risk stratification using routinely available radiotherapy-planning data.
引用
- Wu, Gong, Zhang, et al. Development and validation of a cumulative risk model and assessment tool for radiation-induced hypothyroidism in head and neck cancer: a multicentre study. Clin Transl Radiat Oncol. 2026. PMID: 42959946. doi:10.1016/j.ctro.2026.101325
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。