This study hypothesizes that metabolic characteristics of esophageal tumors can be used to predict treatment response, which considers changes in the primary tumor and lymph nodes, for patients receiving neoadjuvant concurrent chemoradiotherapy (CCRT). This study retrospectively included 60 esophageal cancer patients receiving CCRT followed by surgery. All patients received 18F-fluorodeoxyglucose positron emission tomography/computed tomography (FDG PET/CT) examinations prior to CCRT and in the interval between CCRT and surgery. On the pre-treatment FDG PET/CT images, the maximum standardized uptake value (SUVMaxPre) within the primary tumor was identified. By computerized methods, the CT images of pre- and post-treatment FDG PET/CT were registered. Then, the coordinates of SUVMaxPre were transformed to the post-treatment FDG PET images and delineated a sphere with a diameter of 5 cm to indicate the tumor position. After excluding air, the sphere was partitioned into several metabolic volumes by the optimal dichotomy of high and low metabolic FDG uptakes. Finally, the volume with the shortest distance to the center was adopted and represented by the maximum standardized uptakes (SUVMaxPost). Two additional features, SUVDiff and SUVDiffR, were defined as SUVMaxPost - SUVMaxPre and (SUVMaxPost - SUVMaxPre) / SUVMaxPre. Besides, for defining treatment response, the patients with and without residual tumors were defined as ypT+ and ESOCR based on the histopathology results of surgery. The ESOCR was further classified into pCR to indicate the absence of lymph node metastasis and LNM for remaining. Finally, the area under the receiver operating characteristic curve analysis (AUC) was conducted to assess the features' ability to differentiate two treatment responses. Kruskal-Wallis test was used to evaluate the differences in features between treatment responses. Of the 60 patients, 55 were men (92%), and the mean age was 58. The number of tumors at the esophagus's upper, middle, and lower third were 8, 18, and 34, respectively. Ninety-eight percent of the tumors were squamous cell carcinomas (59/60). The patient numbers of ypT+ and ESOCR were 43 and 17 of which contained 13 pCR and 4 LNM. The SUVDiff and SUVDiffR exhibited a significant ability to identify the ESOCR with AUC = .337 (p = .05) and AUC = .290 (p = .012), respectively. In addition, a statistically significant difference was found among the three groups of ypT+, pCR, and LNM on SUVMaxPre (H = 6.252 and p = .044), SUVDiff (H = 7.948 and p = .019), and SUVDiffR (H = 8.405 and p = .015). In the post-hoc tests corrected by the Bonferroni, the difference between ypT+ and LNM was significant on these features. The metabolic characteristics extracted from pre- and post-treatment FDG PET/CT images could indicate treatment response and disease progression. Further studies are warranted.
Read full abstract