Liu Yiyang, You Yaru, Chen Lihong, Li Jing, Yuan Mengchen, Duan Bo, Ge Huiting, Chen Xuejun, Yang Li, Chen Xingzhi, Li Linfeng, Liu Luhao, Zheng Yue, Li Yiming, Dong Bei, Jia Qian, Ruan Chengwei, Chen Diansen, Hou Zongbin, Zhao Zihao, Ji Qingyu, Gao Jianbo
Department of Radiology, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China.
Henan International Joint Laboratory of Medical Imaging, Zhengzhou, China.
Eur Radiol. 2025 Aug 13. doi: 10.1007/s00330-025-11904-7.
Proliferation marker Ki-67 and immune checkpoint Programmed Cell Death Ligand 1 (PD-L1) are both proteins associated with tumor treatment response. To develop and validate two dynamic nomograms from dual-energy CT (DECT) for noninvasive evaluation of the expression status of these two proteins in advanced gastric cancer (AGC), and explore their associations with treatment response.
In this multicenter prospective and retrospective study, eligible AGC participants who underwent DECT scans were recruited into two study groups (Ki-67 and PD-L1 groups), with their clinical and DECT characteristics collected and measured. The prediction models (Ki-67 dynamic nomogram (KIDN) and PD-L1 dynamic nomogram (PDDN)) were constructed by logistic regression classifier, respectively, and two online calculators were established based on models. The performance of nomograms was comprehensively evaluated.
In total, 459 and 417 eligible patients were analyzed in the Ki-67 and PD-L1 groups, respectively. The nomograms discriminating Ki-67 and PD-L1 achieved AUCs of 0.755 and 0.726, respectively, in the external validation cohort. Additionally, the two nomograms significantly or slightly outperformed any single predictor in both the training and validation cohorts. In both cohorts, KIDN and PDDN showed favorable calibration (Hosmer-Lemeshow test, all p < 0.05) and clinical utility confirmed by decision curve analysis. Moreover, they demonstrated prognostic performance consistent with actual biomarkers and were associated with clinical response to immuno-chemotherapy (both p < 0.05).
The dynamic nomograms, which integrate clinical features and DECT quantitative parameters, enable noninvasive evaluation of treatment-related protein expression in AGC.
Question Noninvasive tools for accurately evaluating Ki-67 and PD-L1 expression status in advanced gastric cancer (AGC) are currently inadequate. Findings Dual-energy CT-based nomograms demonstrated favorable performance in evaluating Ki-67 and PD-L1 expression status and were associated with response to immuno-chemotherapy in AGC. Clinical relevance We developed online calculators based on prediction models that may serve as a supplement to current invasive methods (e.g., immunohistochemistry), potentially aiding therapeutic decision-making.
增殖标志物Ki-67和免疫检查点程序性细胞死亡配体1（PD-L1）均为与肿瘤治疗反应相关的蛋白质。本研究旨在开发并验证基于双能CT（DECT）的两种动态列线图，用于无创评估这两种蛋白质在进展期胃癌（AGC）中的表达状态，并探讨它们与治疗反应的相关性。
在这项多中心前瞻性和回顾性研究中，将接受DECT扫描的符合条件的AGC参与者纳入两个研究组（Ki-67组和PD-L1组），收集并测量其临床和DECT特征。分别通过逻辑回归分类器构建预测模型（Ki-67动态列线图（KIDN）和PD-L1动态列线图（PDDN）），并基于模型建立两个在线计算器。对列线图的性能进行全面评估。
Ki-67组和PD-L1组分别共分析了459例和417例符合条件的患者。在外部验证队列中，区分Ki-67和PD-L1的列线图的曲线下面积（AUC）分别为0.755和0.726。此外，在训练队列和验证队列中，这两种列线图在显著或轻微程度上均优于任何单一预测指标。在两个队列中，KIDN和PDDN均显示出良好的校准（Hosmer-Lemeshow检验，所有p<0.05），决策曲线分析证实了其临床实用性。此外，它们显示出与实际生物标志物一致的预后性能，并与免疫化疗的临床反应相关（均p<0.05）。
整合临床特征和DECT定量参数的动态列线图能够无创评估AGC中与治疗相关的蛋白质表达。
问题：目前用于准确评估进展期胃癌（AGC）中Ki-67和PD-L1表达状态的无创工具不足。发现：基于双能CT的列线图在评估Ki-67和PD-L1表达状态方面表现良好，并与AGC的免疫化疗反应相关。临床意义：我们基于预测模型开发了在线计算器，可作为当前侵入性方法（如免疫组织化学）的补充，可能有助于治疗决策。
