Prediction of Sunitinib Efficacy using Computed Tomography in Patients with Pancreatic Neuroendocrine Tumors (CTpred-Sunitinib-panNET)

Clinically effective methods to predict the efficacy of sunitinib, for patients with metastatic or locally advanced pancreatic neuroendocrine tumors (panNET) are scarce, making precision treatment difficult. This study aimed to develop and validate a computed tomography (CT)-based method to predict the efficacy of sunitinib in patients with panNET.

Pretreatment CT images of 171 lesions from 38 patients with panNET were included. Clinical information including sex, age at diagnosis, progression-free survival of sunitnib treatment, ki-67 index, tumor grade and previous treatment before sunitinib were also collected. CT value ratio (CT value of tumor/CT value of abdominal aorta from the same patient) and radiomics features were extracted for model development. Receiver operating curve (ROC) with area under the curve (AUC) and decision curve analysis (DCA) were used to evaluate the proposed model.

Tumor shrinkage of >10% at first follow-up after sunitinib treatment was significantly associated with longer progression-free survival (PFS; P<0.001) and was used as the major treatment outcome. The CT value ratio could predict tumor shrinkage with AUC of 0.759 (95% confidence interval [CI], 0.685–0.833). We then developed a radiomics signature, which showed significantly higher AUC in training (0.915; 95% CI, 0.866–0.964) and validation (0.770; 95% CI, 0.584–0.956) sets than CT value ratio. DCA also confirmed the clinical utility of the model. Subgroup analysis showed that this radiomics signature had a high accuracy in predicting tumor shrinkage both for primary and metastatic tumors, and for treatment-naive and pretreated tumors. Survival analysis showed that radiomics signature correlated with PFS (P=0.020). The proposed radiomics-based model accurately predicted tumor shrinkage and PFS in patients with panNET receiving sunitinib and may help select patients suitable for sunitinib treatment.

Pancreatic neuroendocrine tumors are a rare group of tumors. The dataset can be used to validate the findings of our study. More importantly, researchers can use this dataset to study the imaging characteristics of pancreatic neuroendocrine tumors.