Our study was aimed to evaluate the prostate grade group concordance of cognitive MR-targeted fusion biopsy with radical prostatectomy and predict the probability of upgrading in a Chinese cohort. For patients diagnosed with PCa via prostate biopsy, the GG of biopsy is critical for physicians to evaluate the condition of cancer and make decisions for the subsequent treatment, especially for patients who may not undergo surgery. Our study also helps clinicians to reduce the probability of underestimation of PCa and establish a more accurate surgery scheme such as intra-fascial prostatectomy and pelvic lymph node dissection.
We successfully constructed a multivariable logistic regression model for accurately predicting the probability of upgrading from prostate biopsy to radical prostatectomy and the final model was acceptable (AUC > 0.7, the Hosmer-Lemeshow test p > 0.05). This result indicated that our model has the potential for improving the accuracy of PCa diagnosis and provide a more precise treatment for patients. We also did nomograms to help clinicians evaluate the risk of upgrading in a specific patient (Fig. 2).
The result of prostate biopsy may not be accurate mainly because of sampling error, heterogeneity of tumor and pathology error. In our study, the GG inconsistency rate from biopsy to radical prostatectomy was 49.7%. In particular, 68.5% of GG1 patients upgrades to higher GGs. Previous studies indicated that 30–50% of GG1 patients may upgrades.2,7,14−16 Though the inconsistency rate was high, the risk factors of upgrading remain controversial: age, PSA, PSAD, prostate volume, number of positive cores and a series of variables were predicted to be independent risk factors in different studies. Our study used the logistic regression model to include and analyze all available variable to achieve a more accurate model predicting upgrading. Provided the 9 variables mentioned in our Method part, our model can calculate the probability of upgrading with high reliability and will be available for providing personalized prognostic information.
There are some deficiencies in our study: a) Our study was a retrospective, single-institution study which means that selection bias was unavoidable and this study type has some inherent disadvantages; b) Some variables, for example, total core percentage and the digital rectal examination result of patient, were not included in our model because these clinical data was missing. Meanwhile, some patients may take 5α deductase inhibitor and some patients may have urinary catheterization. We could not eliminate the influence of these conditions because of the lack of data.