| Objectives The objetive of the curent study was to establishment of prostate cancer Loglstic regression model to assess the possibility of prostate cancer for patients, to improve the diagnosis rate of suspicious prostate cancer, to provide more adequate basis for prostate biopsy.Methods From March 2007 to 2010 March 2010,105 males who underwent transrectal ultrasound prostate biopsy in Xiangya HosPital (PSA≤20ng/ml) were reviewed retrospectively. Prostate cancer was detected in 42 cases. The data set was divided chronologically into a training (model building) set with the first 70 (2/3) men and a test set (model validation) with the 35 subsequent men. Variables analyzed included ages,PSA,PSAD,f/t PSA,Digital rectum Examination,and TRUS fmdings.Logistic regression model was performed to estimate prostate cancer probability using SPSS.Receiver operating characteristic (ROC) curves of the predicted Probability of Logistic regression(P value) were constructed,and relative sensitivity and specificity were calculated.Results Independent predictors of a positive biopsy result included decreased f/t PSA,elevated PSAD,abnormal DRE and abnormal TRUS (P<0.05).Based on the sevariables,a predictive multivariate model was developed. Logit P==-3.427-7.403×f/t PSA+8.019×PSAD+2.064×DRE+1.692×TRUS. ROC of predicted probability of Logistic regression was constructed. Analysis of ROC curve showed that the area under the curve (AUC) was 0.913.The predicted probability of Logistic regression P=0.15 was used to indicate the presense of prostate cancer.The 0.15 cutoff gave a sensitivity of 84.40% and a sepcificity of 75.60%。Conclusions In patients with PSA levels≤20ng/ml,incorporation of f/t PSA,PSAD,DRE and TRUS into Logistic regresion model significantly improved the prediction of prostate cancer and may serve as an aid to reduce unnecesary prostate biopsies.The prostate cancer risk became very high when P value of Logistic regression elevated to more than 0.15. |
PDF Full Text Request