探讨CT特征与现有疗效评价标准相结合对预测接受舒尼替尼治疗的胃肠道间质瘤（GIST）患者进展性疾病（PD）的辅助价值。本次回顾性多中心纳入了81名接受舒尼替尼治疗的GIST患者研究并分为训练组和外部验证组。六个月时的进展被确定为参考标准。比较了 RECIST 1.1 和 Choi 标准的预测性能。分析基线和第一次随访时的 CT 特征。 Logistic 回归分析用于确定最显着的预测因素并制定修改后的标准。81 名患者中，共有 216 个病灶显示出良好的反应，107 个病灶显示出不良反应。 RECIST 1.1 标准在预测进展方面优于 Choi 标准（AUC，0.75 vs. 0.69，p = 0.04）。训练队列中反应良好和反应差的病变之间，扩大/强化的高增强区域、模糊的肿瘤组织界面以及进行性增大的供血或引流肿块的血管 (EVFDM) 存在显着差异（p = 0.001、0.003 和 0.000，分别）。多变量分析显示，扩大/强化的高增强区域 (p = 0.001)、进展性 EVFDM (p = 0.000) 和 RECIST PD (p = 0.000) 是独立的预测因素。制定了修改后的 RECIST (mRECIST) 标准，并显示训练和外部验证队列中的 AUC 显着高于 RECIST 1.1 标准（训练：0.81 与 0.73，p = 0.002；验证：0.82 与 0.77，p = 0.04）。 mRECIST 标准将 CT 特征与 RECIST 1.1 标准相结合，在预测接受舒尼替尼的 GIST 患者的早期进展方面表现出卓越的性能。mRECIST 标准将 CT 特征与 RECIST 1.1 标准相结合，可能有助于早期发现 GIST 患者的进展性疾病。使用舒尼替尼治疗的 GIST 患者，从而有可能指导及时改用晚期药物或联合手术切除。• 在识别舒尼替尼治疗患者的 GIST 进展方面，RECIST 1.1 标准优于 Choi 标准。 • GIST 在 CT 上显示出不同的形态学特征，具体取决于它们对舒尼替尼的反应。 • 将 CT 形态学特征与 RECIST 1.1 标准相结合，可以迅速准确地识别进展性 GIST 病变。© 2023。作者，获得欧洲放射学会的独家许可。

To explore the auxiliary value of combining CT features with existing response evaluation criteria in the prediction of progressive disease (PD) in gastrointestinal stromal tumors (GIST) patients treated with sunitinib.Eighty-one patients with GISTs who received sunitinib were included in this retrospective multicenter study and divided into training and external validation cohorts. Progression at six months was determined as a reference standard. The predictive performance of the RECIST 1.1 and Choi criteria was compared. CT features at baseline and the first follow-up were analyzed. Logistic regression analyses were used to determine the most significant predictors and develop modified criteria.A total of 216 lesions showed a good response and 107 showed a poor response in 81 patients. The RECIST 1.1 criteria performed better than the Choi criteria in predicting progression (AUC, 0.75 vs. 0.69, p = 0.04). The expanded/intensified high-enhancement area, blurred tumor-tissue interface, and progressive enlarged vessels feeding or draining the mass (EVFDM) differed significantly between lesions with good and poor responses in the training cohort (p = 0.001, 0.003, and 0.000, respectively). Multivariate analysis revealed that the expanded/intensified high-enhancement area (p = 0.001), progressive EVFDM (p = 0.000), and RECIST PD (p = 0.000) were independent predictive factors. Modified RECIST (mRECIST) criteria were developed and showed significantly higher AUCs in the training and external validation cohorts than the RECIST 1.1 criteria (training: 0.81 vs. 0.73, p = 0.002; validation: 0.82 vs. 0.77, p = 0.04).The mRECIST criteria, combining CT features with the RECIST 1.1 criteria, demonstrated superior performance in the prediction of early progression in GIST patients receiving sunitinib.The mRECIST criteria, which combine CT features with the RECIST 1.1 criteria, may facilitate the early detection of progressive disease in GIST patients treated with sunitinib, thereby potentially guiding the timely switch to late-line medications or combination with surgical excision.• The RECIST 1.1 criteria outperformed the Choi criteria in identifying progression of GISTs in patients treated with sunitinib. • GISTs displayed different morphologic features on CT depending on how they responded to sunitinib. • Combining CT morphologic features with the RECIST 1.1 criteria allowed for the prompt and accurate identification of progressing GIST lesions.© 2023. The Author(s), under exclusive licence to European Society of Radiology.