| BackgroundNeoadjuvant chemoradiotherapy combined with total mesorectal excision is the standard comprehensive treatment strategy for locally advanced rectal cancer(LARC).TNM(Tumor-Node-Metastasis)staging and tumor regression grade(TRG)is an important reference for predicting the prognosis of LARC patients.However,the individual differences in the efficacy of neoadjuvant therapy make the predictive model currently based on tumor TNM or TRG can not accurately predict the prognosis of LARC patients.The concept of tumor microenvironment(TME)remind that the progress of tumor is the result of the interaction between "seed"(Cancer cell)and "soil"(Stroma),in which collagen fiber is an indispensable component constructing extracellular interstitial skeleton structure and has an important prognostic value.With the development of interdisciplinary research,multi-photon imaging has provided us a powerful technical support for studying collagen alteration in the tumor microenvironment.Therefore,we intend to use multi-photon imaging technology to quantify collagen,explore its relationship with the prognosis of LARC patients,and further construct a prognostic prediction model to assist individualized accurate diagnosis and treatment.ObjectiveAnalyse the collagen alteration in TME quantitatively,explore its association with the prognosis in LARC patients,construct and validate the prognosis predicting model.MethodsThe clinical data of 210 LARC patients who received nCRT and radical surgery from November 2013 to April 2017 in the Fujian Medical University Union Hospital were retrospectively analyzed and as the training cohort.According to the same criteria,155 patients from Nanfang Hospital were collected as independent validation cohort.Multiphoton imaging technique was used to image the pathological tissue and collagen feature extraction.Choosing the valuable collagen feature by the least absolute shrinkage and selection operator(LASSO)COX regression to construct collagen score(Collagen-score).Multivariate Cox regression analysis was used to select the clinicopathological features significantly related to the disease free survival(DFS)and overall survival(OS),which were integrated to construct a nomogram prediction model verified in the validation cohort.ResultsFive collagen characteristics was selected to construct Collagen-score by LASSO COX regression analysis,which were collagen area,collagen fiber number,collagen fiber straightness and two collagen textural features.In the training cohort,the DFS and OS of patients in high Collagen-score group were lower than those in the low Collagen-score group(HR:2.913(95%CI:1.831-4.634),P<0.001;10.669(95%CI:4.667-25.478),P<0.001).Time dependent receiver operating characteristic curve(ROC)detect that Collagen-score has a higher accuracy in predicting the prognosis(The area under ROC(AUROC)of DFS and OS at 1,3 and 5 years were 0.755,0.792,0.746;0.828,0.867 and 0.766 respectively).Univariate Cox regression analysis found that TRG,postoperative adjuvant chemotherapy and Collagen-score were related to DFS,and age,tumor distance from the anus,TRG and Collagen-score were related to OS.Multivariate Cox regression analysis showed that TRG and Collagen-score related to DFS and OS.The clinicopathological factors were age,tumor distance from the anus,TRG and postoperative adjuvant chemotherapy after excluding Collagen-score.These factors were included and constructed Combined Clinicopathological and Collagen nomogram and pure Clinicopathological nomogram models.The predicting acuracy of Combined Clinicopathological and Collagen nomogram model(The AUROC of DFS and OS at 1,3 and 5 years were 0.813,0.826,0.770;0.833,0.902 and 0.787 respectively)better than the pure Clinicopathological nomogram model(The AUROC of DFS and OS at 1,3 and 5 years were 0.722,0.686,0.621;0.723,0.782 and 0.726 respectively).The calibration curve of survival probability shows that the predicted results of Combined Clinicopathological and Collagen nomogram agree well with the actual observation results,and the consistency indexes were 0.779 and 0.868.The decision curve analysis shows that nomograms has certain clinical application value.ConclusionsCollagen-score is an independent predictor of postoperative recurrence,metastasis and death in LARC patients.High Collagen-score patients have poor prognosis.The nomogram prediction model combined with the Collagen-score in TME can improve the prediction ability than traditional clinicopathological features model and help to evaluate LARC patients’ prognosis accurately. |
