Construction Of Prognosis Risk Prediction Model Of Colorectal Cancer With Postoperative Recurrence Or Metastasis Using The Tumor Glucose Metabolism Parameters Of ¹⁸F-FDG PET/CT Imaging And Hematological Inflammatory Markers

Objective:The purpose of the study was to predict the prognosis of patients with colorectal cancer postoperative recurrence or metastasis by investigating the correlation between the tumor glucose metabolism parameters of ¹⁸F-FDG PET/CT imaging and hematological inflammatory markers,determining the combined diagnostic parameters and constructing a nomograms model.Methods:The clinical data of the patients with colorectal cancer postoperative recurrence or metastasis were retrospectively reviewed from August 2015 to September 2020 in our hospital.These patients who were suspected or confirmed by histopathological examination underwent ¹⁸F-FDG PET/CT imaging before treatment.We investigated the correlation among clinical data,histopathological results of the primary tumor,hematological inflammatory markers[neutrophil-to-lymphocyte ratio（NLR）and platelet-to-lymphocyte ratio（PLR）]and the tumor glucose metabolism parameters of ¹⁸F-FDG PET/CT imaging[maximum standardized uptake value（SUVmax）,average standardized uptake value（SUVavg）,peak standardized uptake value（SUVpeak）,maximum standardized uptake value for lean body mass（SULmax）,average standardized uptake value for lean body mass（SULavg）,peak standardized uptake value for lean body mass（SULpeak）,metabolic tumor volume（MTV）and total glycolysis lesion（TLG）].Receiver operating characteristic curve（ROC）was carried out to calculate optimum cut-off point for grouping.Moreover,independent predictors of prognosis were determined and to be used as combined diagnostic parameters using univariate and multivariate COX regression analysis.These the effective prediction biomarkers were used to constructed the predictive model and a nomogram was subsequently drawn.Results:1.Nonparametric tests（Mann-Whitney U test and Kruskal-Wallish H test）were carried out on the correlation between clinical data,histopathological parameters of primary tumor,hematological inflammatory markers and tumor glucose metabolism parameters of ¹⁸F-FDG PET/CT imaging.The results showed:（1）NLR and PLR were significantly higher in rectal cancer patients than that in colon cancer patients.（2）NLR and PLR were gradually decreased with the increasing of the depth of invasion the primary tumors（p T1-2,p T3 and p T4,respectively）.（3）Histopathological findings of primary tumors showed that patients without distant metastasis had higher PLR than those with distant metastasis.（4）The SUVmax,SUVavg,SUVpeak,SULmax SULavg and SULpeak were higher in older age group（>64 years old）than those in younger age group（≤64 years old）.（5）Histopathological findings of primary tumors showed that patients with neural infiltration had lower levels of SULmax,SULavg,MTV,and TLG.（6）TLG increased with the increase of differentiation degree of primary tumor（low,medium and high differentiation）.（7）Histopathological findings of primary tumors showed that patients with cancer embolus had lower MTV and TLG.All of the results above were statistically significant（p<0.05）.2.Spearman correlation analysis showed that there was significant positive correlation between NLR and MTV or TLG（r=0.374 and 0.315,p=0.001 and 0.007）.There was no significant correlation between PLR and SUVmax or SUVavg or SUVpeak or SULmax or SULavg or SULpeak（p>0.05）.3.Univariate and multivariate Cox regression analysis displayed that histopathological findings of the primary tumor with cancer embolus,NLR>2.07 and higher PLR>214.09 were the independent risk factors for overall survival（OS）in colorectal cancer patients with recurrent or metastasis（p<0.01）.4.（1）The area under ROC curve（AUC）of NLR,PLR and SUVmax were 0.712（95%CI:0.613-0.811）,0.615（95%CI:0.508-0.722）and 0.638（95%CI:0.543-0.732）,respectively,while the combined diagnostic parameters of NLR or PLR and SUVmax（CONLR-SUVmax and COPLR-SUVmax）were 0.757（95%CI:0.660-0.853）and 0.692（95%CI:0.583-0.780）,respectively.The results of pair-wise comparison of ROC curves（delong method）showed that there were not only statistically significant differences between the AUC of SUVmax and CONLR-SUVmax（z=-2.798,p=0.0051）but also between the AUC of PLR and COPLR-SUVmax（z=-2.121,p=0.0340）.Therefore,there were no significant difference between the AUC between NLR and CONLR-SUVmax（z=-1.551,p=0.120）or AUC between SUVmax and COPLR-SUVmax（z=-1.213,p=0.2251）.（2）Kaplan-Meier survival analysis showed that CONLR-SUVmax and COPLR-SUVmax were divide into three groups according to their scores（0-2 points）,respectively,and OS among the three groups were significant statistical difference（log-rank method,p<0.001）.5.Age,body mass index（BMI）,p T stage of primary tumor,cancer embolus,CONLR-SUVmax and COPLR-SUVmax were included in the prediction model.Nomogram was drawn subsequently.The model has good discrimination（c-index=0.792）and calibration（bootstrap resampling method）.Conclusions:1.Histopathological findings of the primary tumor with cancer embolus,higher NLR and higher PLR were the independent risk factors for overall survival（OS）of the patients with colorectal cancer with recurrent or metastasis.2.The combined diagnostic parameters（CONLR-SUVmax and COPLR-SUVmax）were more advantages than the single diagnostic parameter（NLR,PLR and SUVmax）in predicting OS of the patients with postoperative recurrence or metastasis of colorectal cancer.3.Nomogram prediction model showed that these factors,including the younger age,lower BMI,lower p T stage of primary tumor,primary tumor located in colon,histopathological findings of the primary tumor with cancer embolus,and high score of CONLR-SUVmax and COPLR-SUVmax,were associated with poor 2-year OS in colorectal cancer patients with recurrent or metastasis.