| Objective:To analyze the correlation between blood test parameters and colorectal cancer,construct a simple and effective prediction model of colorectal cancer based on blood test parameters and validate and analyze the value of the prediction model in clinical prediction of colorectal cancer.Methods:As per case-control study,patients aged 50～75 years were screened who were treated in The First Affiliated Hospital of Dalian Medical University and primarily diagnosed as colorectal cancer by colonoscopy and histopathology from January 1,2019 to December 31,2021 were included in patient group.Persons for whom colorectal cancer was excluded by colonoscopy and histopathology in Physical Examination Center were included in control group.Blood test parameters and the ratios of NLR,PLR,LMR,RLR,PIV,SII,MPV/PLT and PDW/PLT between two groups were compared.All samples were assigned into model group and validation group by the diagnosis time.Multivariate analysis with binary logistic regression was performed for sample data in model group to identify factors of colorectal cancer,construct the multi-factor prediction model and validate the extensionality.The Receiver Operator Characteristic（ROC）curve and the area under ROC curve（AUC）were plotted to evaluate these parameters and calculate the sensitivity and specificity of the model.Results:The mean age of 827 subjects was 62.91±7.20 years.The mean age of patient group（65.11±6.72 years）and control group（59.19±6.43 years）was statistically different（t=12.431,p<0.001）.Patient group included 345 males（66.47%,345/519）and 174 females（33.53%,174/519）,control group included 161 males（52.27%,161/308）and 147 females（47.73%,147/308）,gender distribution was statistically different（χ2=16.414,p<0.001）.Comparison of blood test parameters and ratios between colorectal cancer group and control group:RBC parameters:RBC,HCT,Hb,MCV,MCH and MCHC in colorectal cancer group were lower than those in control group,and RDW-CV and RDW-SD were higher than those in control group,all differences were significant except for RDW-SD（p=0.074）,p<0.001.RBC and WBC parameters:WBC,NEUT,NEUT%,Eos,Eos%,Mono and Mon%in colorectal cancer group were higher than those in control group,and Lymph and Lymph%were lower than those in control group,all differences were significant except for Lymph（p=0.295）,Eos%（p=0.103）and Baso（p=0.646）,p<0.05.PLT-associated parameters:PLT and PCT in colorectal cancer group were higher than those in control group,PDW,MPV and P-LCR were lower than those in control group,all differences were significant except for PCT（p=0.542）and PDW（p=0.587）,p<0.05.Ratios of blood test parameters:NLR,PLR,RLR,PIV and SII in colorectal cancer group were higher than those in control group,LMR was lower than that in control group,the medians of MPV/PLT and PDW/PLT were same between two groups,all differences were significant except for PDW/PLT（p=0.152）,p<0.05.For sample data in model group,gender,age and blood test parameters（n=24,RBC,Hct,RDW-CV,RDW-SD,Hb,MCV,MCH,MCHC,WBC,Neut,Neut%,Lymph,Lymph%,Eos,Eos%,Baso,Baso%,Mono,Mono%,PLT,PCT,P-LCR%,MPV,PDW）were included for analysis by binary logistic stepwise regression（forward LR）.The final function included male（OR=5.361,p<0.001）,age（OR=1.112,p<0.001）,WBC（OR=1.302,p<0.001）and Hb（OR=0.930,p<0.001）as main factors of colorectal cancer.Then 8 ratios of blood test parameters（NLR,PLR,LMR,RLR,Sl I,PIV,MPV/PLT and PDW/PLT）and 24 individual parameters were included for analysis.The results of the function was identical to those without ratios of blood test parameters.The final model function was:Y=1.652+1.679*male（male=1,female=0）+0.106*age+0.264*WBC-0.072*Hb.The prediction probability wasP=1+eyey,and colorectal cancer was indicated when p≥0.5.ROC curve was plotted with the prediction probability obtained by the model of gender,age,WBC and Hb.The AUC of colorectal cancer was 0.837（95%CI:0.804-0.870,p<0.001）calculated by this model,p=0.444（Hosmer-Lemeshow test）,Youden index 0.543.Good discrimination and calibration were seen in this model.AUC values of this model were significantly higher than the AUC values of individual blood test parameters and ratios.Good discrimination and calibration were seen in individual blood test parameters（Hb,Hct,MCHC,WBC,Lymph%and Mono）and ratios of LMR and SII,with AUC>0.600 and p>0.05（Hosmer-Lemeshow test）.The constructed prediction model function was used in validation group and compared with the prediction data of model group:Model group:sensitivity 83.71%,specificity65.50%,total accuracy 77.16%,positive prediction 81.20%and negative prediction69.31%;Validation group:sensitivity 88.34%,specificity 67.59%,total accuracy80.07%,positive prediction 80.45%and negative prediction 79.35%.Both sensitivity and specificity of the model were good validation group.Conclusion:Blood test parameters are important in clinical prediction of colorectal cancer.The prediction model of colorectal cancer has been constructed combining gender,age,WBC and Hb.This model shows good discrimination and calibration,and has been validated with good prediction. |
