Preliminary Study On The Prediction Model For Prostate Cancer

Objective:To improve the diagnosis rate of early prostate cancer,logistic prediction model of prostate cancer was constructed with prostate cancer related indicators.Methods:A total of 235 cases of suspected prostate cancer with prostate biopsy in urology department of our hospital from 2017 to 2020were retrospectively analyzed by three steps.The first step was to analyze the completeness of indicators.From 235 cases various clinical indicators related to prostate cancer were selected,including age(Age),t PSA,f/t PSA,PSAD,prostate volume(PV),digital rectal examination(DRE),MRI signal value(ADC).Univariate analysis was performed to understand the correlation between each single index and the positive rate of prostate biopsy,excluding the cases with incomplete indicators,the remaining 192 cases for the next step.The second step was to analyze the differences of indicators.The data of 192 cases were analyzed by univariate and multivariate analysis to select the indicators(risk factors)with difference,by the stepwise logistic regression analysis five indicators were selected to construct the logistic prediction model of the probability of prostate cancer.The third step was to verify the practicability of the model.In the model validation group,68 cases were highly suspected of prostate cancer due to abnormal PSA,MRI and DRE,but the patients refused prostate biopsy and strongly demanded prostatectomy with the postoperative pathological diagnosis.By the relevant data of 68 patients substituted into the prediction model,the ROC curve was drawn to evaluate the practicability of the model.Results:Univariate and multivariate analysis showed that Age,t PSA,f/t PSA,PSAD,PV,DRE,ADC were related risk factors.After five-step logisitc regression analysis,five indicators for modeling were selected as Age,t PSA,DRE,ADC,PV.The mathematical logic relationship[logit(P)]of these 5 indicators was logit(P)=0.058×Age+0.043×t PSA-4.197×ADC+1.212×DRE-0.016×PV.According to logit(P)the mathematical model for predicting the probability(P)of prostate cancer wasp=1/(1+erp[-logit(P)]),by which the area under the ROC curve of the model validation was 0.869,that indicated the model had better prediction accuracy and was superior to the diagnostic efficiency of all single indicators.Conclusion:The multiparameter logistic prediction model for prostate cancer had a certain practical value for the prediction of early prostate cancer,which was helpful to improve the diagnosis rate of early prostate cancer.