Prediction Of Tumor Immunosuppressive Microenvironment And Outcomes Of Gastric Cancer Based On CT Radiomics

Objective: The tumor immunosuppressive microenvironment can influence treatment response and outcomes.A previously established immunosuppression scoring system(ISS)based on 6 immunosuppressive ligands can be used as an indicator of the immunosuppressive microenvironment and as a robust predictor of prognosis in gastric cancer(GC).However,the current ISS is determined by tissue-based assays performed after surgery,and these assays have many limitations.Thus,the aim of this study was to establish a preoperative Computed tomography(CT)-based radiomic signature of ISS,the radiomic immunosuppressive scoring system(RISS),to predict disease-free survival(DFS)and overall survival(OS)after surgery.Furthermore,we explored whether the RISS could identify patients with stage Ⅱ and Ⅲ GC who might benefit more from postoperative adjuvant chemotherapy.Methods: 1.A total of 525 GC patients who underwent curative resection in Fujian Medical University Union Hospital and Affiliated Hospital of Qinghai University were retrospectively collected and divided into three cohorts: training cohort(n =285),internal validation cohort(n = 116)and external validation cohort 1(n = 116).Radiomic features were extracted from portal venous-phase CT images of GC.A radiomic signature for predicting ISS(RISS)was constructed by using least absolute shrinkage and selection operator(LASSO)regression method.Areas under the curve(AUC)were calculated to evaluate the value of riss in evaluating tumor immunosuppressive microenvironment.2.We also collected GC patients who underwent curative resection in Zhangzhou Affiliated Hospital of Fujian Medical University as external validation cohort 2(n = 116).Kaplan-Meier survival curve and Cox regression analysis were used to evaluate the prognostic value of RISS for GC in 4 cohorts.Nomograms were constructed in combination with other clinicopathological factors.The predictive value of RISS was verified by C-index,calibration curves and decision curve analysis.3.The patients with stage Ⅱ and Ⅲ GC in 4 cohorts were collected.Propensity score matching was performed to minimize potential selection bias and confounding effects between patients who received vs.did not receive chemotherapy using 1:1 nearest matching.Kaplan-Meier survival curve and Cox regression analysis were used to evaluate the relationship between RISS and chemotherapy response.Results: 1.The RISS,which consisted of 10 selected features,showed good discrimination of immunosuppressive status in training cohort(AUC=0.840),internal validation cohort(AUC=0.809),and external validation cohort 1(AUC=0.843).And the AUC values of the RISS were higher than those of any single radiomics feature in 3 cohorts.Based on the receiver operating characteristic(ROC)curve analysis in the training cohort,we classified the patients into a high-RISS group and a low-RISS group with an RISS of 2.15 as the cutoff.2.KM survival curve showed that the prognosis of low-RISS group was significantly better than that of high-RISS group in training cohort,internal validation cohort,external validation cohort 1 and external validation cohort 2(all P<0.05).Multivariate analysis revealed that the RISS was an independent prognostic factor for both disease-free survival(DFS)and overall survival(OS)in all cohorts(all P<0.05),and other prognostic factors included tumor stage and postoperative adjuvant chemotherapy(all P < 0.05).Thus,radiomic nomograms for DFS and OS were built by integrating the RISS and 3 clinicopathologic risk factors,including T stage,N stage,and adjuvant chemotherapy,while nomograms without the RISS were used as clinical models.The radiomic nomograms could more accurately predict OS(Radiomic nomogram vs Clinical model vs TNM stage = 0.748 vs 0.709 vs 0.652)and DFS(Radiomic nomogram vs Clinical model vs TNM stage = 0.743 vs 0.707 vs 0.654)in the training cohort according to the C-index,and similar results were obtained in other 3 cohorts.The decision curve analysis showed that the radiomic nomogram provides a better net benefit for predicting 5-year OS and DFS than the other two models across a range of relevant threshold probabilities in each cohort.3.A total of 529 patients with stage Ⅱ and Ⅲ gastric cancer were included in this study,of which 288 received postoperative adjuvant chemotherapy.After matching,patients with a low RIS derived survival benefit from adjuvant chemotherapy(OS: HR=0.407(95% CI: 0.284-0.584);DFS: HR=0.395(95% CI:0.275-0.568));while those with a high RIS did not(all P>0.05).Conclusions: 1.In this study,a CT-based radiomic signature(RISS)was constructed,which could effectively reflect the immunosuppressive status of GC,so as to realize the noninvasive prediction of the immunosuppressive microenvironment status.2.RISS was closely associated with the prognosis of GC and can be used as an important supplement to the traditional TNM staging to accurately predict the long-term survival of GC after surgery.3.RISS was associated with chemotherapy benefit,which could help identify stage Ⅱ and Ⅲ GC patients most likely to benefit from adjuvant chemotherapy and avoid overtreatment.