Prediction Of Efficacy After Neoadjuvant Chemoradiotherapy For Locally Advanced Rectal Cancer Based On Collagen Score

BackgroundNowadays,neoadjuvant chemoradiotherapy(nCRT)is a standard treatment strategy for locally advanced rectal cancer(LARC).About 20-25%of the patients achieve a pathologic complete response(pCR)after nCRT.Many studies have shown that patients with pCR have longer survival and higher quality of life than patients without pCR.With the deepening of people’s understanding of pCR,local excision or"wait and see" for patients with pCR has been proved to be considered.However,for patients who are insensitivity for nCRT,this treatment cannot bring benefits,at the same time,it will increase the financial burden and cause a series of complications after nCRT for patients.Therefore,how to establish a non-invasive and effective model to safely and accurately predict patients with pCR after nCRT,so as to clinicians enable to selects "tailor-made" treatment according to different patients is a challenge for us.At present,there are many studies about prediction model of pCR,for example clinical pathological data,imageology,gene and more.But there is no research on the microenvironment of tumor.With the development of interdiscipline,multilphoton imaging technology(MIT)provides a powerful tool for us to study the microenvironment of tumor.ObjectiveTo develop and validate a prediction model based on collagen score in the tumor microenvironment to assess the probability of pCR for patients after nCRT.MethodConsecutive patients with LARC by imaging and pathology diagnosis between January 2012 and November 2014 were enrolled in this study with the same inclusion criteria.These patients from Nanfang Hospital of Southern Medical University,the Affiliated Tumor Hospital of Sun Yat-sen University,and the Union Hospital of Fujian Medical University.Finally,a total of 236 patients were included in the study.Then patients were randomized into a training cohort(165 patients)and a validation cohort(71 patients).Collagen features were extracted in each specimen using multiphoton imaging technique,and collagen score was constructed via Lasso regression analysis to select the most predictive parameters in training cohort.After combined with clinicopathological data,binary logistic regression analysis was used to determine the independent predictors of pCR and a prediction model was then constructed and validated in the validation cohort.ResultA 8-paramater-based collagen score via Lasso regression analysis was significantly related to pCR in both the training(P<0.001)and validation coh orts(P<0.001).Subgroup analysis showed that a significant correlation between collagen score and age,gender,BMI,pre-treatment CEA values,pre-treatment T stage,pre-treatment N stage,and biopsy histologic grade.Further analysis revealed that collagen score(OR:14.186,95%CI:4.830-41.666,P<0.001),pre-treatment T stage(OR:4.162,95%CI:1.586-11.025,P=0.004)and pre-treatment CEA values(OR:3.336,95%CI:1.256-8.860,P=0.016)were independent predictors of pCR.A collagen nomogram was then constructed.The area under the ROC for predicting pCR was 0.872(95%CI:0.798-0.947)and 0.849(95%CI:0.752-0.946)in the training and validation cohorts,respectively.The prediction model showed good discrimination and calibration,and decision curves suggested that the collagen nomogram was clinically useful.ConclusionCollagen score based on tumor microenvironment in biopsy tissues is an independent predictor of pCR.The prediction model is useful for decision making in patients with LARC.