Developing a model for forecasting Gleason score ≥7 in potential prostate cancer patients to reduce unnecessary...

The diagnosis of Gleason score (GS) ≥7 with distinction from GS < 7 remains a difficult problem instructing clinical decisions. Moreover, the present wide application of prostate biopsy to increase prostate cancer detection rate might cause unnecessary and excessive examination or treatment. Therefore, a risk assessment model for forecasting GS ≥ 7 in potential prostate cancer patients was established to reduce unnecessary prostate biopsies.

Patients (n = 981; September 2009 to January 2013) who underwent trans-rectal ultrasound (TRUS)-guided core prostate biopsy were retrospectively evaluated in the first stage of the study. Age, prostate-specific antigen (PSA), free PSA (fPSA), the free/total PSA ratio (f/t), prostate volume (PV), PSA density (PSAD), digital rectal examination (DRE) findings (texture, nodules) and B-ultrasound detection results (normal or abnormal, presence of hypoechoic mass or microcalcification) were considered as potential predictive factors. After multiple logistic regression analysis, independent variables used to build a nomogram were selected using a backward elimination selection procedure. Then, a model to forecast GS ≥ 7 was designed for potential prostate cancer patients. In the second stage of the study, 410 cases (January 2013 to March 2015) were subsequently evaluated using our model for prostate biopsies, and the outcomes of biopsies were compared between the two stages.

PSA, DRE texture, DRE nodules and B-ultrasound results were finally brought into our nomogram; a obviously greater area under the receiver operating characteristic (ROC) curve was obtained for the model than utilizing PSA, fPSA or PSAD alone (0.831 vs. 0.803, 0.770, 0.780 separately). We thereafter sought the best cutoff value in the ROC curve at 0.87, which provided sensitivity as high as 90%. Meanwhile, the specificity was 45.8%, which was much higher than the specificity of PSA, fPSA and PSAD at the same sensitivity level (37.7, 24.6 and 35.2%, respectively). In the first stage, the detection rate of GS ≥ 7 in the high-risk group was significantly higher than in the low-risk group (80.3 vs. 35.0%, p < 0.001). Furthermore, in the second stage, with the application of the new model associated with our former models, the rate of GS ≥ 7 was improved from 71.0 (697/981) to 79.2% (267/337) (p = 0.003).

The model for forecasting GS ≥ 7 is effective, which could reduce unnecessary prostate biopsies without delaying patients' diagnoses and treatments.