| [Objective]Head and neck squamous cell carcinoma(HNSCC)is one of the most devastating malignancies with a poor prognosis despite increasingly sophisticated therapeutic approaches.There is growing evidence that epithelial-mesenchymal transition(EMT)-related biomarkers can be used for prognosis prediction in cancer patients.The aim of this study was to develop and validate a reliable prognostic model based on EMT-related gene pairs to predict the prognosis of patients with HNSCC.[Methods]The expression data and corresponding clinicopathological parameters and follow-up data were downloaded from the TCGA database and used as the training dataset.Univariate Cox regression analysis was used to identify EMT-related gene pairs that were significantly associated with overall survival(OS),and then least absolute shrinkage and selection operator(Lasso)analysis was used to screen EMT-related gene pairs for prognostic modeling.All patients with HNSCC were divided into high-and low-risk groups based on the optimal cutoff determined by the receiver operating curve(ROC).Kaplan-Meier(K-M)method was used to generate survival curves,and the Log-rank test was used to estimate the difference between the high-risk and low-risk groups.A separate external validation queue(GSE41613)was used to validate model performance.By comparing the area under the ROC curve(AUC),the performance of our new model and the two models established previously were evaluated.To evaluate the application of this model in the clinical treatment of HNSCC,we evaluated the association between different risk groups with common chemotherapeutic agents and immune checkpoint expression levels.In addition,we integrated clinicopathological parameters and risk scores to establish a nomogram that could individually predict the outcome of patients with HNSCC.[Results]A total of 136 EMT-related gene pairs related to OS were obtained.After LASSO analysis,24 EMT-related gene pairs for model construction were finally obtained.According to the best cut-off value of 0.015,all HNSCC patients were divided into high-risk groups and low-risk groups.The KM curve showed that the mortality rate of the high-risk group was higher than that of the low-risk group,both in the TCGA and GEO data sets(P<0.001).The results of independent prognostic analysis showed that the risk score can be used as an independent predictor to predict the prognosis of HNSCC patients(for TCGA:HR=3.925,95%Cl[2.966-5.193],P<0.001;for GSE41613:HR=2.765,95%CI[1.556-4.914],P<0.001).The prognostic model we developed had the best performance in predicting 3-year OS in HNSCC patients compared to two previously established prognostic models(AUC value=0.804).The levels of common chemotherapeutic agents and immune checkpoint gene expression were statistically different between different risk score groups.By integrating all parameters with independent prognostic significance,we successfully developed nomogram that allow individualized prediction of HNSCC patients.Both the calibration curve and the C-index showed the accuracy of our developed nomogram.Most importantly,the results of the decision curve analysis curves showed that the net benefit of the nomogram was higher than that of the American Joint Committee on Cancer stage system.[Conclusions]In summary,our study used a new algorithm to develop a reliable prognosis model composed of 24 EMT-related gene pairs,which can accurately predict the prognosis of HNSCC patients.The new model helps distinguish those who can benefit from chemotherapy and immunotherapy.In addition,the nomogram established by combining the risk score and other clinicopathological parameters can individually predict the survival rate of patients. |
PDF Full Text Request