A Study On Postoperative Recurrence And Related Factors Of Thoracic Esophageal Squamous Cell Carcinoma

OBJECTIVE: Esophageal cancer is a common malignant tumor of the gastrointestinal tract.Squamous cell carcinoma is the main pathological type.China is a high-risk area of esophageal squamous cell carcinoma（ESCC）.Neoadjuvant therapy plus surgery can improve the survival rate of patients with ESCC,but in China radical surgery is still the main treatment for ESCC.Local recurrence is the main cause of treatment failure in patients with surgical esophageal cancer alone.The median recurrence time is less than one year.Radiotherapy as an important means of postoperative adjuvant therapy can improve the local control rate and improve the prognosis.Radiomics can assess tumor heterogeneity by quantifying medical images and has been shown to correlate with tumor staging,treatment sensitivity.This study intends to start from the patients who have not received adjuvant radiotherapy after ESCC surgery,extract the preoperative CT imaging radiomics parameters,combined with clinical pathological factors to establish individualized prediction of nomogram,to provide assistance for screening patients with high risk of local recurrence.METHODS: Patients who underwent radical resection of ESCC in Shandong Cancer Hospital affiliated to Shandong University from June 2014 to September 2017 and did not receive adjuvant radiotherapy after surgery were retrospectively collected.Intensive CT images within one week prior to surgery,general conditions,intraoperative conditions,postoperative pathological findings,and in-hospital imaging follow-up results were collected.All patients were randomly divided into training and test groups according to 7 to 3.The region of interest was delineated on the arterial phase image using LIFEx4.00 software,and the radiomics parameters were extracted.Based on the training group,LASSO logistic regression was used to select the key features that were significantly associated with local recurrence within one year after surgery,and to construct a radiomics signature.The radiomics signature and the significantly clinical pathological factors were used to establish a predictive model,and the nomogram was drawn.The model is validated in the training and test groups by calibration curves,decision curves,and receiver work curves.RESULTS: A total of 126 patients who met the requirements and had complete follow-up imaging in the hospital were included,including 88 in the training group（77 males,11 females,average age: 60.45 years）,and 38 in the test group（33 males and 5 females,average age: 60.40 years）.68 patients（53.97%）had local recurrence within one year after surgery.LASSO regression screened two characteristic parameters（GLCM_Contrast₃2₂ and GLCM_Contrast₆4₁）that were significantly associated with local recurrence within one year after surgery and constructed a radiomics signature.Chi-square test showed that postoperative pathological stage was significantly associated with local recurrence（P=0.012）.The individualized predictions of image grouping combined with postoperative staging were well predicted in the training（AUC: 0.768,95% CI: 0.670⁰.865）and test groups（AUC: 0.733,95% CI: 0.567⁰.899）.The decision curve showed that the application of nomogram to predict the local recurrence within one year after surgery was better than the separate application of staging and radiomics signature.CONCLUSION: Local recurrence is the main mode of recurrence after radical resection of ESCC.The median recurrence time is less than one year.Pre-treatment contrast-enhanced CT radiomics parameters（GLCM_Contrast₃2₂ and GLCM_Contrast₆4₁）were independent predictors of postoperative local recurrence,and patients with high GLCM_Contrast had a poor prognosis.Pathological staging is an independent predictor of local recurrence after surgery,and patients with late staging have a poor prognosis.The individualized prediction of the use of radiomics signature combined with clinical pathological factors can predict the local recurrence.