Goro Nakayama
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Goro Nakayama has a total of 219 co-authors. Here are the top co-authors by number of shared publications:
- Yasuhiro Kodera
NHO Nagoya Medical Center, Nagoya, Japan
30 shared publications - Masahiko Koike
Nagoya University Graduate School of Medicine
28 shared publications - Michitaka Fujiwara
Nagoya University Graduate School of Medicine, Nagoya, Japan
26 shared publications - Mitsuro Kanda
Department of Gastroenterological Surgery Nagoya University Graduate School of Medicine Nagoya Japan
19 shared publications - Chie Tanaka
Department of Gastroenterological Surgery, Nagoya University Graduate School of Medicine, Nagoya, Japan
18 shared publications - Suguru Yamada
Department of Gastroenterological Surgery, Nagoya Central Hospital, Nagoya, Japan
18 shared publications
Latest Publications
- Open Access
- 19 May 2026
- Published by American Society for Clinical Investigation in Journal of Clinical Investigation
- Vol. 136 (13)
Meflin confers antifibrotic properties to intestinal fibroblasts in inflammatory bowel disease
Dysfunctional intestinal fibrosis is an irreversible complication of Crohn’s disease (CD), The complex heterogeneity of intestinal mesenchymal cells makes it difficult to understand the pathogenesis of intestinal fibrosis. Previously, we identified Meflin as a marker of fibroblast subsets. This study aimed to explore the role of Meflin-positive fibroblasts in intestinal fibrogenesis and investigate the potential of pharmacological control of Meflin expression as a treatment for patients with CD. Our results indicated that Meflin expression was upregulated in fibroblasts at the early stage of fibrosis but was downregulated in established fibrosis in both patients with CD and two different mouse models, which are the chronic dextran sodium sulfate (DSS) model and an interleukin-10-deficient model that spontaneously develops intestinal inflammation. Meflin-deficient mice exacerbated intestinal fibrosis with dysregulated expression of non-canonical Wnt ligand WNT5A and its receptor ROR2. Pharmacologically induced Meflin expression through the administration of a synthetic retinoid reversed intestinal fibrosis in the DSS model and suppressed pro-fibrotic protein secretion in fibroblasts isolated from patients with CD. Our findings indicate that Meflin-positive fibroblasts represent a functional subpopulation that suppresses intestinal fibrosis. Augmentation of Meflin expression shows antifibrotic effects and holds promise as a therapeutic approach for intestinal fibrosis in patients with CD.
- 19 May 2026
- Published by American Society for Clinical Investigation in Journal of Clinical Investigation
- Vol. 136 (13)
Open AccessMeflin confers antifibrotic properties to intestinal fibroblasts in inflammatory bowel disease
Dysfunctional intestinal fibrosis is an irreversible complication of Crohn’s disease (CD), The complex heterogeneity of intestinal mesenchymal cells makes it difficult to understand the pathogenesis of intestinal fibrosis. Previously, we identified Meflin as a marker of fibroblast subsets. This study aimed to explore the role of Meflin-positive fibroblasts in intestinal fibrogenesis and investigate the potential of pharmacological control of Meflin expression as a treatment for patients with CD. Our results indicated that Meflin expression was upregulated in fibroblasts at the early stage of fibrosis but was downregulated in established fibrosis in both patients with CD and two different mouse models, which are the chronic dextran sodium sulfate (DSS) model and an interleukin-10-deficient model that spontaneously develops intestinal inflammation. Meflin-deficient mice exacerbated intestinal fibrosis with dysregulated expression of non-canonical Wnt ligand WNT5A and its receptor ROR2. Pharmacologically induced Meflin expression through the administration of a synthetic retinoid reversed intestinal fibrosis in the DSS model and suppressed pro-fibrotic protein secretion in fibroblasts isolated from patients with CD. Our findings indicate that Meflin-positive fibroblasts represent a functional subpopulation that suppresses intestinal fibrosis. Augmentation of Meflin expression shows antifibrotic effects and holds promise as a therapeutic approach for intestinal fibrosis in patients with CD.
- Open Access
- 8 April 2026
- Published by MDPI AG in Bioengineering
- Vol. 13 (4) , 434
Prognostic Power of Ensemble Learning in Colorectal Cancer with Peritoneal Metastasis: A Multi-Institutional Analysis
Background: Owing to significant clinical heterogeneity, the achievement of accurate survival forecasting for individuals with colorectal cancer and peritoneal metastasis continues to be a complex undertaking. We aimed to transcend traditional prognostic limitations by evaluating machine learning boosting models against standard regression-based methods in terms of estimating overall survival (OS). Methods: Utilizing a multi-institutional registry of 150 patients diagnosed with synchronous peritoneal metastasis of colorectal cancer, we integrated 124 clinicopathological variables to refine our predictive models. Beyond standard preprocessing—including standardization and median imputation—we rigorously compared XGBoost and LightGBM against Ridge, Lasso, and linear regression via five-fold cross-validation. To specifically address right-censoring, an XGBoost Cox model was implemented and validated using Harrell’s C-index, with SHAP and LIME providing essential model interpretability. Results: Boosting models consistently outperformed linear alternatives, which struggled with high error rates and negative R2 values. Specifically, XGBoost achieved an MAE of 475 ± 60 and an RMSE of 585 ± 88. The XGBoost Cox model reached a C-index of 0.64 ± 0.06. SHAP analysis highlighted inflammatory markers and peritoneal disease extent as the most influential prognostic drivers. Conclusions: While boosting models offer a clear accuracy advantage over linear methods, their prognostic power remains moderate. These findings underscore the potential of ensemble learning in oncology, yet mandate external validation before these tools can be integrated into clinical decision-making.
- 8 April 2026
- Published by MDPI AG in Bioengineering
- Vol. 13 (4) , 434
Open AccessPrognostic Power of Ensemble Learning in Colorectal Cancer with Peritoneal Metastasis: A Multi-Institutional Analysis
Background: Owing to significant clinical heterogeneity, the achievement of accurate survival forecasting for individuals with colorectal cancer and peritoneal metastasis continues to be a complex undertaking. We aimed to transcend traditional prognostic limitations by evaluating machine learning boosting models against standard regression-based methods in terms of estimating overall survival (OS). Methods: Utilizing a multi-institutional registry of 150 patients diagnosed with synchronous peritoneal metastasis of colorectal cancer, we integrated 124 clinicopathological variables to refine our predictive models. Beyond standard preprocessing—including standardization and median imputation—we rigorously compared XGBoost and LightGBM against Ridge, Lasso, and linear regression via five-fold cross-validation. To specifically address right-censoring, an XGBoost Cox model was implemented and validated using Harrell’s C-index, with SHAP and LIME providing essential model interpretability. Results: Boosting models consistently outperformed linear alternatives, which struggled with high error rates and negative R2 values. Specifically, XGBoost achieved an MAE of 475 ± 60 and an RMSE of 585 ± 88. The XGBoost Cox model reached a C-index of 0.64 ± 0.06. SHAP analysis highlighted inflammatory markers and peritoneal disease extent as the most influential prognostic drivers. Conclusions: While boosting models offer a clear accuracy advantage over linear methods, their prognostic power remains moderate. These findings underscore the potential of ensemble learning in oncology, yet mandate external validation before these tools can be integrated into clinical decision-making.
- Open Access
- 21 January 2026
- Published by MDPI AG in Preprints.org
Machine Learning-Based Survival Time Prediction in Colorectal Cancer with Peritoneal Metastasis: A Multi-Institutional Registry-Based Study
Background: Prognostic prediction for colorectal cancer patients with peritoneal metastasis remains challenging due to clinical and biological heterogeneity. We aimed to evaluate the utility of machine learning, comparing high-performance boosting models with interpretable regression approaches for overall survival (OS) prediction. Methods: We analyzed a multi-institutional registry cohort of 150 colorectal cancer patients with synchronous peritoneal metastases. A total of 124 variables were included; continuous variables were standardized, categorical variables were one-hot encoded, and missing values were imputed using the median. Models included XGBoost, LightGBM, Ridge regression, Lasso regression, and linear regression. Training was performed with 3-fold cross-validation, and hyperparameters were optimized using Optuna. Evaluation metrics included mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²). Model interpretability was assessed using SHAP values, LIME local explanations, and regression coefficients. Results: Boosting models consistently outperformed linear models. XGBoost achieved the best performance (MAE 424, RMSE 526, R² = 0.04), while LightGBM showed comparable accuracy. In contrast, Ridge, Lasso, and Linear regression yielded high errors (MAE > 900, RMSE > 1200) with negative R² values, indicating poor predictive ability. SHAP analysis highlighted systemic inflammation markers (CRP, BUN), surgical assessment of tumor depth, operative factors (time, bleeding), and peritoneal metastasis characteristics as major determinants of OS. LIME analyses further provided case-specific interpretability, identifying feature contributions in long-, intermediate-, and short-term survivors. Conclusion: Boosting models, particularly XGBoost, demonstrated superior performance compared with traditional regression models in predicting OS for colorectal cancer patients with peritoneal metastasis, although absolute predictive accuracy remains modest. Integration of SHAP and LIME linked model outputs with clinically plausible prognostic factors, enhancing interpretability. Ensemble learning may provide a promising adjunct for prognostic assessment and should be validated in larger, genomically enriched cohorts.
- 21 January 2026
- Published by MDPI AG in Preprints.org
Open AccessMachine Learning-Based Survival Time Prediction in Colorectal Cancer with Peritoneal Metastasis: A Multi-Institutional Registry-Based Study
Background: Prognostic prediction for colorectal cancer patients with peritoneal metastasis remains challenging due to clinical and biological heterogeneity. We aimed to evaluate the utility of machine learning, comparing high-performance boosting models with interpretable regression approaches for overall survival (OS) prediction. Methods: We analyzed a multi-institutional registry cohort of 150 colorectal cancer patients with synchronous peritoneal metastases. A total of 124 variables were included; continuous variables were standardized, categorical variables were one-hot encoded, and missing values were imputed using the median. Models included XGBoost, LightGBM, Ridge regression, Lasso regression, and linear regression. Training was performed with 3-fold cross-validation, and hyperparameters were optimized using Optuna. Evaluation metrics included mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²). Model interpretability was assessed using SHAP values, LIME local explanations, and regression coefficients. Results: Boosting models consistently outperformed linear models. XGBoost achieved the best performance (MAE 424, RMSE 526, R² = 0.04), while LightGBM showed comparable accuracy. In contrast, Ridge, Lasso, and Linear regression yielded high errors (MAE > 900, RMSE > 1200) with negative R² values, indicating poor predictive ability. SHAP analysis highlighted systemic inflammation markers (CRP, BUN), surgical assessment of tumor depth, operative factors (time, bleeding), and peritoneal metastasis characteristics as major determinants of OS. LIME analyses further provided case-specific interpretability, identifying feature contributions in long-, intermediate-, and short-term survivors. Conclusion: Boosting models, particularly XGBoost, demonstrated superior performance compared with traditional regression models in predicting OS for colorectal cancer patients with peritoneal metastasis, although absolute predictive accuracy remains modest. Integration of SHAP and LIME linked model outputs with clinically plausible prognostic factors, enhancing interpretability. Ensemble learning may provide a promising adjunct for prognostic assessment and should be validated in larger, genomically enriched cohorts.
- Open Access
BRAF V600E and Non-V600E Mutations in RAS Wild-Type Metastatic Colorectal Cancer: Prognostic and Therapeutic Insights from a Nationwide, Multicenter, Observational Study (J-BROS)
Background/Objectives: BRAF mutations occur in 5–10% of metastatic colorectal cancer (mCRC) cases, but their implications for prognosis and optimal treatment remain unclear. Methods: This multicenter, prospective observational study analyzed 377 RAS wild-type cases from 511 patients across 32 centers, using PCR-based methods. Results: BRAF mutations were identified in 21% (79/377) of cases, predominantly V600E (89.9%) with a minority of non-V600E (10.1%). Microsatellite instability (MSI) testing revealed MSI-high in 11.3%, exclusively among V600E cases. V600E mutations were linked to right-sided tumors, poor differentiation, and elevated CA19-9 levels. Median survival was significantly lower in V600E cases compared to BRAF wild-type (12.4 vs. 37.5 months, HR 3.25, p < 0.001) and marginally lower non-V600E cases (12.4 vs. 34.7 months, HR 0.61, p = 0.057). Chemotherapy regimens (doublet vs. triplet) and targeted treatments (bevacizumab vs. anti-EGFR) showed no significant survival differences in V600E patients. Similarly, RAS/BRAF wild-type patients had comparable survival with bevacizumab versus anti-EGFR, even for left-sided tumors. Conclusions: These findings highlight distinct clinical and prognostic profiles for BRAF V600E and non-V600E mutations, while treatment choice appears to have limited impact on survival in these subgroups or RAS/BRAF wild-type cases.
Open AccessBRAF V600E and Non-V600E Mutations in RAS Wild-Type Metastatic Colorectal Cancer: Prognostic and Therapeutic Insights from a Nationwide, Multicenter, Observational Study (J-BROS)
Background/Objectives: BRAF mutations occur in 5–10% of metastatic colorectal cancer (mCRC) cases, but their implications for prognosis and optimal treatment remain unclear. Methods: This multicenter, prospective observational study analyzed 377 RAS wild-type cases from 511 patients across 32 centers, using PCR-based methods. Results: BRAF mutations were identified in 21% (79/377) of cases, predominantly V600E (89.9%) with a minority of non-V600E (10.1%). Microsatellite instability (MSI) testing revealed MSI-high in 11.3%, exclusively among V600E cases. V600E mutations were linked to right-sided tumors, poor differentiation, and elevated CA19-9 levels. Median survival was significantly lower in V600E cases compared to BRAF wild-type (12.4 vs. 37.5 months, HR 3.25, p < 0.001) and marginally lower non-V600E cases (12.4 vs. 34.7 months, HR 0.61, p = 0.057). Chemotherapy regimens (doublet vs. triplet) and targeted treatments (bevacizumab vs. anti-EGFR) showed no significant survival differences in V600E patients. Similarly, RAS/BRAF wild-type patients had comparable survival with bevacizumab versus anti-EGFR, even for left-sided tumors. Conclusions: These findings highlight distinct clinical and prognostic profiles for BRAF V600E and non-V600E mutations, while treatment choice appears to have limited impact on survival in these subgroups or RAS/BRAF wild-type cases.
- 29 February 2024
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 31 (5) , 1-2
ASO Visual Abstract: Absence of Hypercoagulation Status After Neoadjuvant Treatment is Associated with Favorable Prognosis in Patients Undergoing Subtotal Esophagectomy for Esophageal Squamous Cell Carcinoma
- 29 February 2024
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 31 (5) , 1-2
ASO Visual Abstract: Absence of Hypercoagulation Status After Neoadjuvant Treatment is Associated with Favorable Prognosis in Patients Undergoing Subtotal Esophagectomy for Esophageal Squamous Cell Carcinoma
- 20 January 2024
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 31 (5) , 3417-3425
Absence of Hypercoagulation Status after Neoadjuvant Treatment is Associated with Favorable Prognosis in Patients Undergoing Subtotal Esophagectomy for Esophageal Squamous Cell Carcinoma
Abnormal activation of the coagulation system is associated with malignant tumor progression. Although neoadjuvant treatment (NAT) for resectable esophagea
- 20 January 2024
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 31 (5) , 3417-3425
Absence of Hypercoagulation Status after Neoadjuvant Treatment is Associated with Favorable Prognosis in Patients Undergoing Subtotal Esophagectomy for Esophageal Squamous Cell Carcinoma
Abnormal activation of the coagulation system is associated with malignant tumor progression. Although neoadjuvant treatment (NAT) for resectable esophagea
- Open Access
G protein subunit gamma 4 expression has potential of detection, prediction, and therapeutic target for liver metastasis of gastric cancer
Liver metastasis of gastric cancer is the most common for hematogenous metastases and so fatal, that the identification of novel markers and targets for therapy are crucial. We conducted transcriptome analyses between synchronous liver metastasis, primary tumor, and adjacent tissues from four patients with metastasis confined to the liver to discover that GNG4 upregulated substantially in primary gastric cancer tissues. Quantitative RT-qPCR assay for 300 gastric cancer patients revealed that higher levels of GNG4 in primary cancer were associated with shorter overall survival and a higher risk of liver recurrence. The oncogenic phenotypes of GNG4 were determined by knockout and forced expression of GNG4 . Tumor formation by GNG4 knockout cells was more strikingly attenuated in a liver metastasis mouse model compared with a subcutaneous model. GNG4 is a candidate for a therapeutic target for liver metastasis, and its expression may enable us to provide better disease monitoring for liver metastasis.
Open AccessG protein subunit gamma 4 expression has potential of detection, prediction, and therapeutic target for liver metastasis of gastric cancer
Liver metastasis of gastric cancer is the most common for hematogenous metastases and so fatal, that the identification of novel markers and targets for therapy are crucial. We conducted transcriptome analyses between synchronous liver metastasis, primary tumor, and adjacent tissues from four patients with metastasis confined to the liver to discover that GNG4 upregulated substantially in primary gastric cancer tissues. Quantitative RT-qPCR assay for 300 gastric cancer patients revealed that higher levels of GNG4 in primary cancer were associated with shorter overall survival and a higher risk of liver recurrence. The oncogenic phenotypes of GNG4 were determined by knockout and forced expression of GNG4 . Tumor formation by GNG4 knockout cells was more strikingly attenuated in a liver metastasis mouse model compared with a subcutaneous model. GNG4 is a candidate for a therapeutic target for liver metastasis, and its expression may enable us to provide better disease monitoring for liver metastasis.
- Open Access
- 15 October 2019
- Published by Springer Nature in BMC Surgery
- Vol. 19 (1) , 1-8
Feasibility of subtotal esophagectomy with systematic lymphadenectomy in selected elderly patients with esophageal cancer; a propensity score matching analysis
Background: The global increase in elderly populations is accompanied by an increasing number of candidates for esophagectomy. Here we aimed to determine the postoperative outcomes after subtotal esophagectomy in elderly patients with esophageal cancer. Methods: Patients (n = 432) with who underwent curative-intent transthoracic subtotal esophagectomy with 2- or 3-field lymphadenectomies for thoracic esophageal cancer were classified as follows: non-elderly (age < 75 years, n = 373) and elderly (age ≥ 75 years, n = 59) and groups. To balance the essential variables including neoadjuvant treatment and stage of progression, we conducted propensity score analysis, and clinical characteristics, perioperative course and prognosis were compared. Results: After two-to-one propensity score matching, 100 and 50 patients were classified in the non-elderly and elderly groups. The elderly group had more comorbidities and lower preoperative cholinesterase activities and prognostic nutrition indexes. Although incidences of postoperative pneumonia, arrhythmia and delirium were slightly increased in the elderly group, no significant differences were observed in overall incidence of postoperative complications, rates of repeat surgery and death caused by surgery, and length of postoperative hospital stay between the two groups. There were no significant differences in disease-free and disease-specific survival as well as overall survival between the two groups. Conclusion: Older age (≥75 years) had limited impact on morbidity, disease recurrence, and survival after subtotal esophagectomy. Therefore, age should not prevent older patients from benefitting from surgery.
- 15 October 2019
- Published by Springer Nature in BMC Surgery
- Vol. 19 (1) , 1-8
Open AccessFeasibility of subtotal esophagectomy with systematic lymphadenectomy in selected elderly patients with esophageal cancer; a propensity score matching analysis
Background: The global increase in elderly populations is accompanied by an increasing number of candidates for esophagectomy. Here we aimed to determine the postoperative outcomes after subtotal esophagectomy in elderly patients with esophageal cancer. Methods: Patients (n = 432) with who underwent curative-intent transthoracic subtotal esophagectomy with 2- or 3-field lymphadenectomies for thoracic esophageal cancer were classified as follows: non-elderly (age < 75 years, n = 373) and elderly (age ≥ 75 years, n = 59) and groups. To balance the essential variables including neoadjuvant treatment and stage of progression, we conducted propensity score analysis, and clinical characteristics, perioperative course and prognosis were compared. Results: After two-to-one propensity score matching, 100 and 50 patients were classified in the non-elderly and elderly groups. The elderly group had more comorbidities and lower preoperative cholinesterase activities and prognostic nutrition indexes. Although incidences of postoperative pneumonia, arrhythmia and delirium were slightly increased in the elderly group, no significant differences were observed in overall incidence of postoperative complications, rates of repeat surgery and death caused by surgery, and length of postoperative hospital stay between the two groups. There were no significant differences in disease-free and disease-specific survival as well as overall survival between the two groups. Conclusion: Older age (≥75 years) had limited impact on morbidity, disease recurrence, and survival after subtotal esophagectomy. Therefore, age should not prevent older patients from benefitting from surgery.
- Top Cited PapersOpen Access
- 3 January 2019
- Published by Springer Nature in Journal of Translational Medicine
- Vol. 17 (1) , 1
Identification of a serum-based miRNA signature for response of esophageal squamous cell carcinoma to neoadjuvant chemotherapy
Neoadjuvant chemotherapy (NAC) has become the standard of care for resectable esophageal squamous cell carcinoma (ESCC) which is one of the most lethal cancers, to improve resectability and prognosis. On this basis, to provide individually optimized therapy for ESCC, a minimally-invasive biomarker for response to NAC is strongly desired. This study aimed to identify the miRNA signature in serum specimens taken from ESCC patients undergoing NAC through genome-wide microarray technology. Comprehensive miRNA-expression profiles of serum specimens from ESCC patients before initial treatment were analyzed using microarray. A qPCR assay was performed to test the robustness of identified serum-based miRNA signature for discriminating response to NAC with serum specimens taken from 100 ESCC cases undergoing NAC. We prioritized 62 miRNAs differentially expressed between responders and non-responders (absolute log2 fold change > 1.0, corresponding P < 0.05) and from the 62 miRNAs, we selected the miR-23a-5p, miR-193b-5p, and miR-873-3p, which were highly expressed in non-responders. Following qPCR analysis indicated the expression of miR-193b-5p and miR-873-3p in serum specimens were significantly higher in non-responders among three selected miRNAs (P = 0.004 and 0.001, respectively). Subsequently, we developed 2-miR-model (miR-193b-5p and miR-873-3p), 3-miR-model, and 2-miR + lymphatic invasion (ly) model based on logistic regression analysis, which achieved the better area under the receiver operating characteristic curves than those of single miRNAs as 2-miR-model, 0.70 (95% CI 0.57 to 0.82); 3-miR-model, 0.70 (95% CI 0.57 to 0.83); and 2-miR + ly, 0.73 (95% CI 0.60–0.86), respectively. Furthermore, we compared the detective power of the combined model: 2-miR + ly for discriminating non-responders to NAC, to other pretreatment clinical features. Consequently, 2-miR + ly model was superior to serum SCC antigen with great significance (P = 0.01) and to ly, and clinical T stage with marginal significance (P = 0.18, 0.07, respectively). Collectively, we demonstrated that the potential of a multi-miRNA biomarker for identifying NAC response in ESCC is realistic, and can be used in the clinic with the further validation.
- 3 January 2019
- Published by Springer Nature in Journal of Translational Medicine
- Vol. 17 (1) , 1
Top Cited PapersOpen AccessIdentification of a serum-based miRNA signature for response of esophageal squamous cell carcinoma to neoadjuvant chemotherapy
Neoadjuvant chemotherapy (NAC) has become the standard of care for resectable esophageal squamous cell carcinoma (ESCC) which is one of the most lethal cancers, to improve resectability and prognosis. On this basis, to provide individually optimized therapy for ESCC, a minimally-invasive biomarker for response to NAC is strongly desired. This study aimed to identify the miRNA signature in serum specimens taken from ESCC patients undergoing NAC through genome-wide microarray technology. Comprehensive miRNA-expression profiles of serum specimens from ESCC patients before initial treatment were analyzed using microarray. A qPCR assay was performed to test the robustness of identified serum-based miRNA signature for discriminating response to NAC with serum specimens taken from 100 ESCC cases undergoing NAC. We prioritized 62 miRNAs differentially expressed between responders and non-responders (absolute log2 fold change > 1.0, corresponding P < 0.05) and from the 62 miRNAs, we selected the miR-23a-5p, miR-193b-5p, and miR-873-3p, which were highly expressed in non-responders. Following qPCR analysis indicated the expression of miR-193b-5p and miR-873-3p in serum specimens were significantly higher in non-responders among three selected miRNAs (P = 0.004 and 0.001, respectively). Subsequently, we developed 2-miR-model (miR-193b-5p and miR-873-3p), 3-miR-model, and 2-miR + lymphatic invasion (ly) model based on logistic regression analysis, which achieved the better area under the receiver operating characteristic curves than those of single miRNAs as 2-miR-model, 0.70 (95% CI 0.57 to 0.82); 3-miR-model, 0.70 (95% CI 0.57 to 0.83); and 2-miR + ly, 0.73 (95% CI 0.60–0.86), respectively. Furthermore, we compared the detective power of the combined model: 2-miR + ly for discriminating non-responders to NAC, to other pretreatment clinical features. Consequently, 2-miR + ly model was superior to serum SCC antigen with great significance (P = 0.01) and to ly, and clinical T stage with marginal significance (P = 0.18, 0.07, respectively). Collectively, we demonstrated that the potential of a multi-miRNA biomarker for identifying NAC response in ESCC is realistic, and can be used in the clinic with the further validation.