Longitudinal CT-based deep learning radiomics for predicting prognosis in esophageal squamous cell carcinoma treated with definitive chemoradiotherapy: a two-center study.

Mean platelet volume (MPV), as a marker of platelet activity, has been shown to be an efficient prognostic biomarker in several types of cancer. Using MPV, this study aimed to create and validate a prognostic nomogram to the overall survival in esophageal squamous cell carcinoma (ESCC) patients. The nomogram was constructed and tested using data from a retrospective study of 1893 patients who were randomly assigned to the training and testing cohorts with a 7:3 randomization. In order to screen out the optimal predictors for overall survival (OS), we conducted the LASSO-cox regression, univariate, and multivariate cox regression analyses. Subsequently, the predictive accuracy of the nomogram was validated in both the training and the testing cohorts. Finally, decision curve analysis (DCA) was used to confirm clinical validity. Age, MPV, nerve invasion, T stage, and N stage were found as independent prognostic variables for OS and were further developed into a nomogram. The nomogram's prediction accuracy for 1-, 3-, and 5-year OS was 0.736, 0.749, 0.774, and 0.724, 0.719, 0.704 in the training and testing cohorts, respectively. Furthermore, DCA results indicated that nomograms outperformed the AJCC 8th and conventional T, N staging systems in both the training and testing cohorts. The nomogram, in conjunction with MPV and standard clinicopathological markers, could improve the accuracy of prediction of OS in ESCC patients.

Excision repair cross-complementing group 1 (ERCC1) functions as a nucleotide excision repair (NER) enzyme. Altered ERCC1 expression or function is closely associated with cancer development and progression. This study determined the association of ERCC1 expression with survivin expression, clinicopathological characteristics, and survival of esophageal squamous cell carcinoma (ESCC) patients after postoperative concurrent chemoradiotherapy.Tissue specimens from 102 resected ESCC patients were acquired for immunohistochemical analysis of ERCC1 and survivin protein expression.ERCC1 expression was detected in 62.7% of ESCC tissues and in 9.8% of normal squamous epithelium tissues (P < .01), while survivin expression was detected in 60.8% of ESCC tissues and in 19.6% of normal squamous epithelia (P < .01). ERCC1 overexpression associated with advanced tumor clinical stage and lymph node metastasis (P < .05), but not with tumor size, depth of invasion, or differentiation (P > .05). ERCC1 overexpression was also associated with survivin levels (r = 0.42, P < .01) and worse progression-free survival of ESCC patients after concurrent chemoradiotherapy. Multivariate analysis data revealed that ERCC1 and survivin protein expression were independent predictors of overall survival of ESCC patients after chemotherapy and/or radiotherapy (P < .05).ERCC1 overexpression is an important phenotype that is associated with ESCC lymph node metastasis and advanced tumor clinical stages. ERCC1 expression may also inhibit ESCC cell apoptosis via regulating survivin expression, and ERCC1 and survivin overexpression are independent predictors of prognosis for ESCC patients who receive chemotherapy and/or radiotherapy.

Purpose Various malnutrition and inflammation criteria were associated with prognosis of esophageal squamous cell carcinoma (ESCC) patients. Nonetheless, the interplay of clinicopathological features, malnutrition, and inflammation criteria with overall survival in ESCC patients remains unclear. Methods We retrospectively reviewed medical records of 205 patients diagnosed with ESCC between 2007 and 2012, and evaluated the status of participant malnutrition and inflammation, including body mass index < 18.5 kg/m2, body weight loss > 5.0%, serum albumin level < 3.5 g/dl, neutrophil-to-lymphocyte ratio > 3.5, platelet-to-lymphocyte ratio > 20, prognostic nutrition index < 40, blood total lymphocyte count < 1600 cells/mm3, and grades of body mass index-adjusted body weight loss (combined BMI-BWL). We assessed the association of clinicopathological features, nutritional status, and inflammation condition with overall survival using univariate and multivariate Cox regression analyses. Results The mean overall survival of ESCC patients was 28.8 mo,. The multivariate logistic regression model after adjustment for clinicopathological variables, malnutrition status, inflammation condition, and co-morbid status found that tumor stage and grades of combined BMI-BML served as equally important prognostic factors for overall survival. Conclusions Advanced tumor stage and high grades of combined BMI-BWL were independent prognostic factors for overall survival in ESCC patients.

The purpose of this study was to investigate the impact of clinicopathological factors on the overall survival (OS) of advanced esophageal squamous cell carcinoma (ESCC) patients with both lymph node and distant metastasis and build a nomogram for OS prediction. We selected 621 ESCC patients with cT1-4N1-3M1 stage without surgical treatment from the Surveillance, Epidemiology, and End Results (SEER) database and randomized (in a 7:3 ratio) to the training cohort and internal validation cohort. Another 159 patients were enrolled from the Cancer Hospital of Shantou University Medical College as the external validation cohort. A nomogram was developed based on independent risk factors that resulted from a multivariate Cox regression analysis. Receiver operating characteristic (ROC) curves and the area under the ROC curve (AUC) were used to evaluate the discriminative ability and calibration curves were constructed to evaluate the calibration ability. Kaplan-Meier survival analysis and log-rank tests were then used to predict the further OS status of these patients. The multivariate Cox regression analysis revealed that sex, T stage, radiotherapy, and chemotherapy were independent prognostic factors for ESCC patients with cT1-4N1-3M1 stage. All these factors were incorporated to construct a nomogram. The prognostic nomogram in training cohort exhibited the AUCs of 0.784, 0.746, and 0.735 for predicting 6-, 9-, and 12-month OS, respectively. Calibration curves exhibited that the nomogram-predicted OS were insistent with the actual OS. In validation cohorts, the nomogram still showed acceptable discrimination ability and calibration. All individuals were allocated into high-risk versus low-risk groups based on the median risk score of the training cohort. The OS of the high-risk group was shorter than that of the low-risk group in three cohorts. We developed and validated an individualized survival prediction nomogram for predicting OS in ESCC patients with cT1-4N1-3M1 stage, which may help clinicians to assess the situation of advanced ESCC patients and implement further treatment.

Simple SummaryThe prognosis of esophageal squamous cell carcinoma (ESCC) patients is poor, with a five-year survival of 15–34%. We examined the expression of STAT3α and STAT3β in pretreatment tumor biopsies of 105 ESCC patients who received concurrent chemoradiotherapy (CCRT) by immunohistochemistry. The data showed that ESCC patients who demonstrate both high STAT3α expression and high STAT3β expression in the cytoplasm have a significantly better survival rate. Moreover, the ESCC patients with high STAT3β expression have a complete response to concurrent chemoradiotherapy. STAT3β-overexpressed ESCC cell lines exhibit CCRT (platinum plus radiation therapy) sensitivity, resulting in cell death. RNA sequencing found that ESCC cells highly expressing STAT3β undergo necrosis after CCRT. In summary, STAT3β could be potentially used to predict the response to CCRT, which may provide an important insight into the treatment of ESCC.Concurrent chemoradiotherapy (CCRT), especially platinum plus radiotherapy, is considered to be one of the most promising treatment modalities for patients with advanced esophageal cancer. STAT3β regulates specific target genes and inhibits the process of tumorigenesis and development. It is also a good prognostic marker and a potential marker for response to adjuvant chemoradiotherapy (ACRT). We aimed to investigate the relationship between STAT3β and CCRT. We examined the expression of STAT3α and STAT3β in pretreatment tumor biopsies of 105 ESCC patients who received CCRT by immunohistochemistry. The data showed that ESCC patients who demonstrate both high STAT3α expression and high STAT3β expression in the cytoplasm have a significantly better survival rate, and STAT3β expression is an independent protective factor (HR = 0.424, p = 0.003). Meanwhile, ESCC patients with high STAT3β expression demonstrated a complete response to CCRT in 65 patients who received platinum plus radiation therapy (p = 0.014). In ESCC cells, high STAT3β expression significantly inhibits the ability of colony formation and cell proliferation, suggesting that STAT3β enhances sensitivity to CCRT (platinum plus radiation therapy). Mechanistically, through RNA-seq analysis, we found that the TNF signaling pathway and necrotic cell death pathway were significantly upregulated in highly expressed STAT3β cells after CCRT treatment. Overall, our study highlights that STAT3β could potentially be used to predict the response to platinum plus radiation therapy, which may provide an important insight into the treatment of ESCC.

Lymphocyte-to-monocyte ratio, neutrophil-to-lymphocyte ratio, and platelet-to-lymphocyte ratio represent systemic immune-inflammatory responses. We evaluated the association between immune-inflammatory cell ratios and prognosis in esophageal squamous cell carcinoma (ESCC) patients who underwent definitive concurrent chemoradiotherapy (dCCRT). Medical records of 68 ESCC patients in three institutions who underwent dCCRT between 2006 and 2017 were reviewed. The immune-inflammatory cell ratios were calculated before and after dCCRT. The median follow-up time was 11.4 months. The 3-year overall survival (OS) rate was 21.6%. Among the immune-inflammatory cell ratios, lower post-dCCRT neutrophil-to-lymphocyte ratio (NLRpost) was associated with better OS (median 15.2 vs. 9.7 months, p=0.030). Patients with lower NLRpost had more improved OS when adjuvant chemotherapy was administered following dCCRT (median 16.6 vs. 4.8 months, p<0.001). NLRpost may be useful in predicting OS in ESCC patients after dCCRT. Furthermore, NLRpost might play a role in establishing adjuvant therapy plans following dCCRT.

BackgroundPrognosis of esophageal squamous cell carcinoma (ESCC) patients is poor and the concurrent chemoradiation therapy (CCRT) provided to ESCC patients often failed due to resistance. Therefore, development of biomarkers for predicting CCRT response is immensely important. In this study, we evaluated the predicting value of SRY (sex determining region Y)-box 17 (SOX17) protein during CCRT and its dysregulation of transcriptional targets in CCRT resistance in ESCC.MethodsPyrosequencing methylation, RT-qPCR and immunohistochemistry assays were performed to examine the DNA methylation, mRNA expression and protein expression levels of SOX17 in endoscopic biopsy from a total of 70 ESCC patients received CCRT. Cell proliferation, clonogenic survival and xenograft growth were used to confirm the sensitization of ESCC cell line KYSE510 in response to cisplatin, radiation or CCRT treatment by SOX17 overexpression in vitro and in vivo. Luciferase activity, RT-qPCR and ChIP-qPCR assays were conducted to examine transcription regulation of SOX17 in KYSE510 parental, KYSE510 radio-resistant cells and their derived xenografts.ResultsHigh DNA methylation coincided with low mRNA and protein expression levels of SOX17 in pre-treatment endoscopic biopsy from ESCC patients with poor CCRT response. SOX17 protein expression exhibited a good prediction performance in discriminating poor CCRT responders from good responder. Overexpression of SOX17 sensitized KYSE510 radio-resistant cells to cisplatin, radiation or CCRT treatment in cell and xenograft models. Importantly, SOX17 transcriptionally down-regulated DNA repair and damage response-related genes including BRCA1, BRCA2, RAD51, KU80 DNAPK, p21, SIRT1, NFAT5 and REV3L in KYSE510 radio-resistant cells to achieve the sensitization effect to anti-cancer treatment. Low expression of BRCA1, DNAPK, p21, RAD51 and SIRT1 was confirmed in SOX17 sensitized xenograft tissues derived from radio-resistant ESCC cells.ConclusionsOur study reveals a novel mechanism by which SOX17 transcriptionally inactivates DNA repair and damage response-related genes to sensitize ESCC cell or xenograft to CCRT treatment. In addition, we establish a proof-of-concept CCRT prediction biomarker using SOX17 immunohistochemical staining in pre-treatment endoscopic biopsies to identify ESCC patients who are at high risk of CCRT failure and need intensive care.

Background: The findings of a recent analysis on microRNAs (miRNAs) suggest that circulating miRNAs have potential as biomarkers of esophageal squamous cell carcinoma (ESCC). In order to identify specific miRNAs of ESCC, we analyzed the circulating miRNAs of patients who underwent endoscopic mucosal resection (EMR) and esophagectomy. Method: After obtaining written informed consent, we collected paired (pre and post treatment) blood samples from 60 superficial ESCC patients and 42 pretreatment advanced ESCC patients between 2011 and 2015. Samples were divided into training (40 superficial ESCC and 22 advanced ESCC) and test (15 superficial ESCC and 20 advanced ESCC) cohorts according to the period at which they were obtained (between 2011 and 2013 and between 2014 and 2015). Fifty-five patients underwent EMR and were confirmed as stage 0 or Stage 1a. Microarray analyses of blood samples were performed using the 3D-Gene miRNA microarray platform (Toray). Normalization was achieved using the Quantile method, and poor quality samples were excluded from the analysis. Any two clinical groups were compared using a two-sided Student’s t-test. miRNAs exhibiting significant differences were subsequently evaluated using a logistic regression analysis (LRA). Multivariate LRA, Akaike’s Information Criterion (AIC), and Receiver Operating Characteristic (ROC) analyses were performed in order to evaluate the diagnostic power of miRNA combinations. In all series, we used post-EMR patients as control cases. Results: Twelve miRNAs (miR-6722-5p, 489, 4525, 409-3p, 6088, 3678-5p, 197-5p, 4281, 5090, 3173-3p, 762, and 1470) were selected as discriminant markers (S-combination: AUC 1.00) of superficial ESCC, while 4 miRNAs (miR-4723-3p, 4646-3p, 2392, and 1236-3p) were selected as discriminant markers (A-combination: AUC 1.00) of advanced ESCC in the training cohort. There were no overlap miRNAs between the two combinations. In the test cohort, the S-combination discriminated superficial ESCC (AUC 1.00), while the A-combination discriminated advanced ESCC (AUC 1.00) from post-EMR patients. Furthermore, the S-combination discriminated advanced ESCC in the test cohort (AUC 1.00). However, the A-combination did not clearly discriminate superficial ESCC (AUC 0.833). Conclusion: Our results suggest that selected miRNAs are useful biomarkers for the discrimination of ESCC. However, biomarkers of superficial ESCC and advanced ESCC may differ. Citation Format: Yutaka Shimada, Yoshinori Takei, Tomoyuki Okumura, Takuya Nagata, Haruka Fujinami, Miwako Arima, Tetsuya Abe, Yasumasa Niwa, Masahiro Tajika, Tetsuo Sudo, Kazuharu Shimizu. Circulating microRNA expression profiles as a novel diagnostic biomarker for esophageal squamous cell carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4430. doi:10.1158/1538-7445.AM2017-4430

BackgroundEarly detection of esophageal squamous cell carcinoma (ESCC) recurrence is a key element for follow‐up care and surveillance. The aim of this study is to detect the level of circulating exosomes (CEs) in ESCC patient and clarify its clinical significance.MethodsIn this study, 200 serum samples of ESCC patients were obtained from the Zhejiang Cancer Hospital Biospecimen Repository. Total CEs were purified by selectively capturing epithelial cell adhesion molecule positive exosomes, using magnetic‐bead technique. enzyme‐linked immunosorbent assay (ELISA) was performed to measure the concentration level of CEs. The oncogenic potential of CEs was analyzed in vitro.ResultsSerum concentration of CEs was significantly higher in ESCC patients than in healthy controls (P < 0.01). Receiver‐operating characteristic curve analysis demonstrated that CEs concentration could distinguish patients with ESCC from healthy individuals with a sensitivity of 75% and a specificity of 85%. Kaplan‐Meier analysis demonstrated that the increased CEs concentration was associated with poor overall survival (P = 0.01) and progression free survival (P = 0.03) in ESCC patients. Multivariate cox regression analysis revealed that CEs concentration was an independent prognostic marker for overall survival in ESCC patients (P < 0.01). Results from transwell and wound scratching experiments showed that the CEs could promote cell migration and invasion.ConclusionsThis study clearly demonstrates that CEs from ESCC patients are stable enough to be measured and their levels in ESCC patients are significantly upregulated. Circulating exosomes could serve as a novel noninvasive biomarker for detection of ESCC. Their involvement in carcinogenesis must be further established.

PurposeThis study is aimed to develop and validate a machine learning model, which combined radiomics and clinical characteristics to predicting the definitive chemoradiotherapy (dCRT) treatment response in esophageal squamous cell carcinoma (ESCC) patients. Methods: 204 advanced ESCC patients were included who underwent dCRT at our hospital. Patients were randomly divided into training cohort and validation cohort with a ratio of 7:3. The radiomics features were selected by LASSO algorithm. The clinical features were selected by multivariate logistics analysis (p < 0.05). Subsequently, a combined radiomics and clinical model was established and validated to predict the treatment response in ESCC patients by logistic regression model. The performance of the model was evaluated by receiver operating characteristic (ROC) curve, decision curve analysis (DCA), nomogram, and calibration curve. Results: Total of 944 radiomics features were extracted from the pre-treatment contrasted enhanced CT images (CECT). After feature selection, 3 radiomics features and 3 clinical features were identified as the most predictive variables. The combined model shows better prediction performance among radiomics model or clinical model. The radiomics model’s AUC values in training and validation cohort are 0.71,0.69. As for clinical model the AUC values were 0.74,0.75 in training and validation cohort. However, the AUC values in combined model are 0.79, 0.78 in training cohort and validation cohort, respectively. DCA and calibration curve also demonstrated good performance for the combined model. Conclusion: The radiomics combined clinical features model demonstrates superior treatment response prediction ability for ESCC patients received dCRT. This model has the potential to assist clinicians in identifying non-responsive patients before treatment and guide individualized therapy for advanced ESCC patients.

BackgroundClinical T4 stage (cT4) esophageal tumors are difficult to be surgically resected, and definitive radiotherapy (RT) or chemoradiotherapy (dCRT) remains the main treatment. The study aims to analyze the association between the status of lymph node (LN) metastasis and survival outcomes in the cT4 stage esophageal squamous cell carcinoma (ESCC) patients that underwent treatment with dCRT or RT.MethodsThis retrospective study analyzed the clinical data of 555 ESCC patients treated with dCRT or RT at the Shandong Cancer Hospital and the Liaocheng People’s Hospital from 2010 to 2017. Kaplan–Meier and Cox regression analyses was performed to determine the relationship between LN metastasis and survival outcomes of cT4 and non-cT4 ESCC patients. The chi-square test was used to evaluate the differences in the local and distal recurrence patterns in the ESCC patients belonging to various clinical T stages.ResultsThe 3-year survival rates for patients with non-cT4 ESCC and cT4 ESCC were 47.9% and 30.8%, respectively. The overall survival (OS) and progression-free survival (PFS) rates were strongly associated with the status of LN metastasis in the entire cohort (all P < 0.001) and the non-cT4 group (all P < 0.001) but not in the cT4 group. The local recurrence rates were 60.7% for the cT4 ESCC patients and 45.1% for the non-cT4 ESCC patients (P < 0.001). Multivariate analysis showed that clinical N stage (P = 0.002), LN size (P = 0.007), and abdominal LN involvement (P = 0.011) were independent predictors of favorable OS in the non-cT4 group. However, clinical N stage (P = 0.824), LN size (P = 0.383), and abdominal LN involvement (P = 0.337) did not show any significant correlation with OS in the cT4 ESCC patients.ConclusionsOur data demonstrated that the status of LN metastasis did not correlate with OS in the cT4 ESCC patients that received dCRT or RT. Furthermore, the prevalence of local recurrence was higher in the cT4 ESCC patients.

Predicting the neoadjuvant chemoradiotherapy (NCRT) treatment response in esophageal squamous cell carcinoma (ESCC) patients remains challenging. This study aims to evaluate the value of CT imaging-based machine learning models for predicting pathologic complete response (pCR) in ESCC patients receiving NCRT and to establish correlations with their underlying biology via radiogenomics analysis. Methods We identified 231 eligible patients from two centers. Handcrafted radiomics features (analyzed by PyRadiomics) and deep learning features (analyzed by transfer learning using the convolutional neural network, Xception) were extracted from pretreatment CT images. A handcrafted radiomics model and deep learning model were built with support vector machine. The models were trained in the training cohort (n = 161) and validated in an external testing cohort (n = 70). The radiological model with better performance was incorporated into a nomogram. Pathway enrichment analysis in a subset of the cohort (n = 28) with gene expression profiles revealed the potential biological processes correlated with the radiological prediction. Results We constructed a seven-feature handcrafted radiomics model and an eight-feature deep learning model to predict NCRT response. The deep learning model outperformed its counterpart (AUC: 0.76 vs. 0.73; accuracy: 71.4% vs. 65.7%) and was used to establish a nomogram model incorporating cN staging, which showed good discrimination in the testing cohort (C-index = 0.76, accuracy = 72.9%). Radiogenomics analysis illustrated a potential association between the radiological prediction and underlying molecular processes involving multiple factors, including WNT and TGF- signaling pathways, the microenvironment, radiation response and mitotic nuclear division. Conclusion We developed a CT imaging-based machine learning model that showed satisfactory performance in NCRT response prediction for ESCC patients. Further radiogenomics analysis provided useful insights into the biological mechanisms of therapy resistance in ESCC. These findings could enhance the applications of radiological characteristics in precision oncology and clinical practice.

Abstract: Objective Malnutrition and cachexia are common in esophageal squamous cell carcinoma (ESCC) patients undergoing radiotherapy. This study evaluated how malnutrition- and cachexia-related indicators, including the albumin-to-globulin ratio (AGR), and their changes during radiotherapy predict the treatment outcomes. Methods We reviewed a total of 172 ESCC patients receiving radiotherapy (as a primary cohort) and performed a Cox regression analysis on potential prognostic factors, including the AGR, as well as the TNM stage and concomitant chemotherapy. A subsequent receiver operating characteristics (ROC) curve and Kaplan Meier survival analysis was performed for ESCC patients stratified by the average AGR cut-off point. Results In addition to the well-documented factors (i.e. TNM stage and concomitant chemotherapy), the average AGR was a significant prognostic factor for ESCC patients receiving radiotherapy. By plotting the ROC curve of the average AGR with regard to the ESCC prognosis, we obtained cut-off points for the overall patients (cut-off point: 1.5, AUC: 0.636), for patients with stage II/III disease (cut-off point: 1.5, AUC: 0.611), and for patients with stage IV disease (cut-off point: 1.84, AUC: 0.900). The ESCC patients with higher average AGR had a significantly more favorable OS compared with those with a lower average AGR. Notably, ESCC patients who had an increasing ΔAGR during radiotherapy had a considerably more favorable OS compared with those with a decreasing ΔAGR. All of these findings were reproducible in the validation cohort. Conclusion A lower average AGR is indicative of a poorer prognosis for ESCC patients following radiotherapy. Improving the nutritional status and preventing or ameliorating cachexia might contribute to improving the prognosis of ESCC patients.

Preoperative prediction of lymph node (LN) metastasis is accepted as a crucial independent risk factor for treatment decision-making for esophageal squamous cell carcinoma (ESCC) patients. Our study aimed to establish a non-invasive nomogram to identify LN metastasis preoperatively in ESCC patients. Construction of the nomogram involved three sequential phases with independent patient cohorts. In the discovery phase (N = 20), LN metastasis-associated microRNAs (miRNAs) were selected from next-generation sequencing (NGS) assay of human ESCC serum exosome samples. In the training phase (N = 178), a nomogram that incorporated exosomal miRNA model and clinicopathologic was developed by multivariate logistic regression analysis to preoperatively predict LN status. In the validation phase (n = 188), we validated the predicted nomogram's calibration, discrimination, and clinical usefulness. Four differently expressed miRNAs (chr 8-23234-3p, chr 1-17695-5p, chr 8-2743-5p, and miR-432-5p) were tested and selected in the serum exosome samples from ESCC patients who have or do not have LN metastasis. Subsequently, an optimized four-exosomal miRNA model was constructed and validated in the clinical samples, which could effectively identify ESCC patients with LN metastasis, and was significantly superior to preoperative computed tomography (CT) report. In addition, a clinical nomogram consisting of the four-exosomal miRNA model and CT report was established in training cohort, which showed high predictive value in both training and validation cohorts [area under the receiver operating characteristic curve (AUC): 0.880 and 0.869, respectively]. The Hosmer–Lemeshow test and decision curve analysis implied the nomogram's clinical applicability. Our novel non-invasive nomogram is a robust prediction tool with promising clinical potential for preoperative LN metastasis prediction of ESCC patients, especially in T1 stage.

PurposeTo develop a nomogram model for predicting local progress-free survival (LPFS) in esophageal squamous cell carcinoma (ESCC) patients treated with concurrent chemo-radiotherapy (CCRT).MethodsWe collected the clinical data of ESCC patients treated with CCRT in our hospital. Eligible patients were randomly divided into training cohort and validation cohort. The least absolute shrinkage and selection operator (LASSO) with COX regression was performed to select optimal radiomic features to calculate Rad-score for predicting LPFS in the training cohort. The univariate and multivariate analyses were performed to identify the predictive clinical factors for developing a nomogram model. The C-index was used to assess the performance of the predictive model and calibration curve was used to evaluate the accuracy.ResultsA total of 221 ESCC patients were included in our study, with 155 patients in training cohort and 66 patients in validation cohort. Seventeen radiomic features were selected by LASSO COX regression analysis to calculate Rad-score for predicting LPFS. The patients with a Rad-score ≥ 0.1411 had high risk of local recurrence, and those with a Rad-score < 0.1411 had low risk of local recurrence. Multivariate analysis showed that N stage, CR status and Rad-score were independent predictive factors for LPFS. A nomogram model was built based on the result of multivariate analysis. The C-index of the nomogram was 0.745 (95% CI 0.7700–0.790) in training cohort and 0.723(95% CI 0.654–0.791) in validation cohort. The 3-year LPFS rate predicted by the nomogram model was highly consistent with the actual 3-year LPFS rate both in the training cohort and the validation cohort.ConclusionWe developed and validated a prediction model based on radiomic features and clinical factors, which can be used to predict LPFS of patients after CCRT. This model is conducive to identifying the patients with ESCC benefited more from CCRT.