文献速递头颈部肿瘤放疗放疗物理与技术

头颈癌质子治疗中预测模型性能指标与患者选择结果的关系

The relation between prediction model performance measures and patient selection outcomes for proton therapy in head and neck cancer.
PubMed 文献 · Radiother Oncol · 2022-12-22

在头颈癌患者数据中模拟正常组织并发症概率(NTCP)模型验证与患者选择过程,通过扰动模型独立改变AUC、整体校准(CITL)和校准斜率(CS),观察其与ΔNTCP及单模型患者选择差异的关系。AUC小幅降低(-0.02)对应ΔNTCP均值变化0.9-3.2%、单模型患者选择差异2-19%;CITL或CS偏离(-0.2或+0.2)对应ΔNTCP均值变化0.3-1.4%、单模型患者选择差异1-10%。作者建议考虑选择AUC最高的模型,必要时重新校准。

要点

  • AUC降低0.02对应ΔNTCP均值变化0.9-3.2%、单模型患者选择差异2-19%。
  • CITL或CS偏离±0.2对应ΔNTCP均值变化0.3-1.4%、单模型患者选择差异1-10%。
  • 每个性能指标独立影响ΔNTCP和患者选择,应在有代表性且足够大的外部样本中评估。

核心结论

  • 每个模型性能指标独立影响ΔNTCP和患者选择,应在有代表性且足够大的外部样本中评估。
  • 作者建议的实用模型选择方法是考虑AUC最高的模型,必要时对其进行重新校准。

研究设计与样本

  • 在实证头颈癌患者数据中模拟模型验证及随后的患者选择过程。
  • 通过模型扰动独立操纵性能指标,研究模型性能与患者选择之间的关系。

主要发现

  • AUC小幅降低(-0.02)产生ΔNTCP均值变化0.9-3.2%,单模型患者选择差异2-19%。
  • CITL或CS偏离(-0.2或+0.2)产生ΔNTCP均值变化0.3-1.4%,单模型患者选择差异1-10%。

关键数值与推荐

  • 推荐的模型评价指标包括受试者工作特征曲线下面积(AUC)、整体校准(CITL)和校准斜率(CS)。
  • AUC降低0.02对应ΔNTCP均值变化0.9-3.2%、单模型患者选择差异2-19%。
  • CITL或CS偏离±0.2对应ΔNTCP均值变化0.3-1.4%、单模型患者选择差异1-10%。

临床怎么用

  • 作者建议的实用模型选择方法是考虑AUC最高的模型,必要时对其进行重新校准。

局限性

  • 原文未报告局限性

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: Normal-tissue complication probability (NTCP) models predict complication risk in patients receiving radiotherapy, considering radiation dose to healthy tissues, and are used to select patients for proton therapy, based on their expected reduction in risk after proton therapy versus photon radiotherapy (ΔNTCP). Recommended model evaluation measures include area under the receiver operating characteristic curve (AUC), overall calibration (CITL), and calibration slope (CS), whose precise relation to patient selection is still unclear. We investigated how each measure relates to patient selection outcomes. METHODS: The model validation and consequent patient selection process was simulated within empirical head and neck cancer patient data. By manipulating performance measures independently via model perturbations, the relation between model performance and patient selection was studied. RESULTS: Small reductions in AUC (-0.02) yielded mean changes in ΔNTCP between 0.9-3.2 %, and single-model patient selection differences between 2-19 %. Deviations (-0.2 or +0.2) in CITL or CS yielded mean changes in ΔNTCP between 0.3-1.4 %, and single-model patient selection differences between 1-10 %. CONCLUSIONS: Each measure independently impacts ΔNTCP and patient selection and should thus be assessed in a representative sufficiently large external sample. Our suggested practical model selection approach is considering the model with the highest AUC, and recalibrating it if needed.

来源:Radiother Oncol,查看原文。版权归原作者所有。

引用

  1. Leeuwenberg AM, Reitsma JB, Van den Bosch LGLJ, et al. The relation between prediction model performance measures and patient selection outcomes for proton therapy in head and neck cancer. Radiother Oncol. 2022. PMID: 36566991. doi:10.1016/j.radonc.2022.109449

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。