Combining the radiomic features and traditional parameters of ¹⁸F-FDG PET with clinical profiles to improve prognostic stratification in patients with esophageal squamous cell carcinoma treated with neoadjuvant chemoradiotherapy and surgery

Objectives: To investigate the role of the traditional and radiomic parameters of ¹⁸F-FDG PET for predicting the outcomes of patients with esophageal squamous cell carcinoma (SqCC). Methods: Forty-four patients with primary esophageal SqCC who underwent neoadjuvant chemoradiotherapy (CCRT) followed by esophagectomy (tri-modality treatment) were retrospectively analyzed. All patients underwent ¹⁸F-FDG PET/CT before and after neoadjuvant CCRT. The radiomic features were calculated using the pre-treatment PET scan. Pre-treatment radiomic features and changes in the PET-derived traditional parameters after neoadjuvant CCRT were analyzed according to the pathological response to esophagectomy, disease-free survival (DFS), and overall survival (OS). We further developed a scoring system based on the independent survival prognosticators and compared our model to the traditional TNM staging system and surgical pathology. Results: A pre-treatment primary tumor histogram entropy ≥ 3.69 predicts an unfavorable response to neoadjuvant CCRT (OR = 19.25, p = 0.009). An SUVmax reduction ratio ≤ 0.76, a pre-treatment primary tumor code similarity ≤ 0.0235, and incomplete pathological remission were independently associated with poor OS (p = 0.019, 0.033, and 0.038, respectively) and DFS (p = 0.049, 0.021, and 0.009, respectively). The three survival prognosticators were used to construct a scoring system (score 0–1, 2, and 3). Patients with a score of 2 or 3 had a significantly worse survival outcome than those with a score of 0–1 (HRs for OS: 3.58 for score 2, and 15.19 for score 3, p < 0.001; HRs for DFS: 1.39 for score 2 and 6.04 for score 3, p = 0.001).This survival prediction model was superior to the traditional TNM staging system (p < 0.001 versus p = 0.061 for OS, and p = 0.001 versus p = 0.027 for DFS) and the model based on surgical pathology (p < 0.001 versus p = 0.049 for OS, and p = 0.001 versus p = 0.022 for DFS). Conclusions: The ¹⁸F-FDG PET-derived radiomic parameter is useful for predicting the surgical pathological response in patients with esophageal SqCC treated with the tri-modality method. Using a combination of traditional and radiomic PET parameters with clinical profiles enables better stratification of patients into subgroups with various survival rates.