The use of prostate specific antigen density to predict clinically significant prostate cancer

The purpose of this study was to assess the predictive value of prostate specific antigen density (PSAD) for detection of clinically significant prostate cancer in men undergoing systematic transrectal ultrasound (TRUS)-guided prostate biopsy. We retrospectively analyzed data of men who underwent TRUS-guided prostate biopsy because of elevated PSA (≤ 20 ng/ml) or abnormal digital rectal examination. Receiver operating characteristic curve analysis to compare PSA and PSAD performance and chi-square automatic interaction detector methodologies were used to identify predictors of clinically significant cancer (Gleason score ≥ 7 or international society of urological pathology grade group ≥ 2). Nine-hundred and ninety-two consecutive men with a median age of 66 years (IQR 61–71) were included in the study. Median PSAD was 0.10 ng/ml2 (IQR 0.10–0.22). Prostate adenocarcinoma was diagnosed in 338 men (34%). Clinically significant prostate adenocarcinoma was diagnosed in 167 patients (50% of all cancers and 17% of the whole cohort). The AUC to predict clinically significant prostate cancer was 0.64 for PSA and 0.78 for PSAD (P < 0.001). The highest Youden's index for PSAD was at 0.20 ng/ml2 with 70% sensitivity and 79% specificity for the diagnosis of clinically significant cancer. Men with PSAD < 0.09 ng/ml2 had only 4% chance of having clinically significant disease. The detection rate of clinically significant prostate cancer in patients with PSAD between 0.09 and 0.19 ng/ml2 was significantly higher when prostate volume was less than 33 ml. In conclusion, PSAD was a better predictor than PSA alone of clinically significant prostate cancer in patients undergoing TRUS-guided biopsy. Patients with PSAD below 0.09 ng/ml2 were unlikely to harbor clinically significant prostate cancer. Combining PSAD in the gray zone (0.09–0.19) with prostate volume below 33 ml adds diagnostic value of clinically significant prostate cancer.


Patients and methods
Patients. Soroka University Medical Center ethics committee approved the study and waived informed consent requirements and all methods were performed in accordance with the relevant guidelines and regulations. We retrospectively collected data on 1,226 consecutive men who underwent TRUS-guided prostate biopsy at our center between January 2014 and December 2018. Men with PSA levels of > 4 ng/ml, or with abnormal DRE findings, were referred for TRUS-guided prostate biopsy. Ideally, prostate biopsies were performed with 12 cores taken from both lobes, 6 from each side, under TRUS guidance; two specimens were obtained from the base of the prostate gland of each side, two from middle of the gland, and two from the apex. Occasionally, more cores were taken when suspicious lesion was noticed on ultrasonography. All biopsies were performed only by experienced senior urologists. Since repeat biopsies for the same patients were included in the database, we selected the first available biopsy for each patient performed at our center and we excluded all repeated biopsies. Biopsy specimens were reviewed by experienced pathologists. Our cohort included 992 men with the restriction of PSA ≤ 20 ng/ml measured prior to biopsy. Estimated prostate volume was measured by TRUS. The PSAD (ng/ ml 2 ) was calculated as PSA (ng/mL) divided by the prostate volume (ml).

Statistical analysis.
Descriptive statistics of the study sample were used to summarize relevant participant characteristics. The Student's t-test was used for comparison of the continuous variables. Chi-square and Fisher's exact tests were used for comparison of the proportions. Receiver operating characteristic (ROC) curves were plotted as sensitivity vs. 1-specificity for PSA and PSAD for each outcome. Clinically significant prostate cancer is defined as Gleason score ≥ 7 or international society of urological pathology (ISUP) grade group 2 or greater. We calculated the PSA and PSAD AUC predicting clinically significant disease and overall prostate cancer vs. no cancer 13 . We have used Youden's index (sensitivity + specificity-1) for identification of the optimum cut-off point for PSAD to be used as a predictor of the significant finding. Classification and regression tree and χ2 automatic interaction detection (CHAID) methods were used to divide the predictors into categories on the basis of the clinically significant cancer detection status 14 . All tests were two-tailed and statistical significance was defined as a P < 0.05. Analyses were performed using SPSS Statistics for Windows, version 24 (IBM Corp., Armonk, N.Y., USA).

Results
Nine-hundred ninety-two consecutive men with a median age of 66 years (IQR 61-71) who had TRUS-guided prostate biopsy were included in the study. Median PSA was 6.2 ng/ml (IQR 5-8.3) and median prostate volume was 43 ml (IQR 31-59). Median PSAD was 0.10 ng/ml 2 (IQR 0.10-0.22). Prostate adenocarcinoma was diagnosed in 338 men (34%). Table 1 summarizes the clinical and histopathological features of the men with cancer vs. men without cancer. Men with prostate cancer were slightly older (median age of 68 vs. 66) with significantly higher mean PSA and PSAD values but lower mean prostate volume. Of the men with cancer about 50% had clinically significant prostate cancer, ISUP grade group ≥ 2 (Gleason score ≥ 7).
We next used the CHAID methodology to create a decision tree and found that patients with PSAD higher than 0.34 have 56.4% chance to be diagnosed with clinically significant cancer while patients with PSAD less than 0.09 have very low probability (4%) of having clinically significant prostate cancer ( Table 2). Men with PSAD between 0.19 to 0.34 have 31.5% chance to be diagnosed with clinically significant prostate cancer. The chance of having clinically significant prostate cancer in men with PSAD between 0.09 and 0.19 is relatively low (8.5%). In this subgroup of patients, when PSAD is combined with prostate volume, the detection rate increases to 16.5% specifically in men having prostate volume less or equal to 33 ml while it drops to 6.8% when prostate volume is larger than 33 ml ( Table 2).

Discussion
In the era of post PSA screening there is still great concern about the low reliability of PSA to detect clinically significant prostate cancer. The use of PSAD was first developed by Benson et al. to improve cancer detection rate when PSA ranged between 4 and 10 ng/ml (gray zone) 15 . Later on, Catalona et al. found high cancer detection performance (95% sensitivity) when they used a PSAD cutoff of 0.078 ng/ml 2 in men with PSA values in this gray zone 16 . They found that 40% of cancer cases would have been missed with a cutoff of 0.15 ng/ml 2 in this population. It should be taken into consideration that PSAD value was originally based on the sextant biopsy template described by Bazinet et al. 17 , which is not the standard anymore. It is clear that the lower the threshold is set, the more biopsies are done and fewer cancers are missed but the rate of unnecessary biopsies and over detection of non-clinically significant disease are greater.
Recently, Aminsharifi et al. studied the predictive accuracy of PSAD for the detection of clinically significant prostate cancer 8 . Interestingly, they found that using a PSAD cutoff of 0.08 ng/ml 2 could have avoided 273 of 2,162 biopsies (13%), missing 48 of 622 non-clinically significant prostate cancers (ISUP grade group 1) (8%) and www.nature.com/scientificreports/ www.nature.com/scientificreports/ 10 of 499 clinically significant cancers (2%). Furthermore, Nordström et al. reported that using PSAD cutoffs of 0.1 and 0.15 ng/ml 2 could lead to missing detection rate of 23% and 51%, respectively, for clinically significant cancers 9 . But, when they set a cutoff point of 0.07 ng/ml 2 , they were able to avoid about 20% of biopsies while missing only about 7% of clinically significant cancers in patients with PSA 3 ng/ml or greater. It should be kept in mind that these recent studies are based on a higher number of biopsies from the prostate compared to the classic sextant biopsy.
In the current study, the performance of PSA was relatively poor compared with PSAD. Also in our population a cutoff point at PSAD of 0.15 gave the highest sensitivity (70%) and specificity (70%) for the detection of any prostate cancer, while at PSAD of 0.20 gave the highest sensitivity (70%) and specificity (79%) for the detection of clinically significant cancer (Fig. 1).
In addition, we used the CHAID methodology to create a decision tree and found that patients with PSAD higher than 0.19 have 30% to 60% chance to be diagnosed with clinically significant cancer while patients with PSAD less than 0.09 have very low probability (4%) of having clinically significant prostate cancer ( Table 2). Interestingly, the chance of having clinically significant prostate cancer in men with PSAD in a "gray zone" between 0.09 and 0.19 is relatively low (8.5%). In this subgroup of patients falling in the PSAD gray zone, the detection rate of clinically significant caner depends on prostate volume. When prostate volume is higher than 33 ml the detection rate is 6.8% whereas it increases more than two-fold to (17%) when prostate volume is smaller than 33 ml ( Table 2). The chance of hitting a cancerous lesion is higher in smaller prostates than in larger ones . In our analysis, this was true and significant only in a subgroup of patients with PSAD values between 0.09 and 0.19 indicating that we are not dealing with an artifact. Otherwise, it would have be observed in each value of PSAD. Table 2. CHAID decision tree for the detection of clinically significant prostate cancer. *, "Others" consists of patients without cancer and non-clinically significant cancer. **, percent relates to total number of patients (992). CHAID, chi-square automatic interaction detector; PSAD, prostate-specific antigen density; CS, clinically-significant. www.nature.com/scientificreports/ Our findings are consistent with those recently described by Jue et al. who evaluated the predictive accuracy of PSAD to detect prostate cancer in 1,290 men with various PSA levels 10 . It should be emphasized that our cohort included patients with PSA levels ≤ 20 ng/ml.
The tendency of PSAD to detect clinically significant prostate cancer has been shown in several clinical scenarios. For instance, the integration of PSAD into active surveillance protocols can be associated with improved enrollment criteria and a reduced rate of upgrading and reclassification down to 17.5% 11,12 . It is worth to mention that today there are additional commercially available blood tests to help diagnose clinically significant prostate cancer better than PSA such as 4Kscore and prostate health index (PHI) 18 as well as urine-based tests including PCA-3 19 and SelectMDx 20 .
The limitations of this study that it includes data specific to one population from one area located in southern Israel. Therefore, results may not be extrapolated completely to other populations. Second, this is an observational retrospective study lacking information on the indications for PSA testing which limits our conclusions since no prospective data was collected. Third, the results should be interpreted with caution, since prostate volume was determined only after referral of the patients to TRUS-guided prostate biopsy, but not for the purpose of calculating prostate volume before the decision on the biopsy had been made. Forth, it should be emphasized that only PSA and PSAD were used as predictive variables. Finally, we only included individuals with the first biopsy available without data on repeated biopsies. Therefore, performance and the cutoff point may not be generalizable to candidates for repeat biopsy although PSAD was shown to be a significant predictive factor for positive repeat biopsy 21 .
In summary, PSAD was a better predictor than PSA alone of biopsy outcomes in patients undergoing transrectal prostate biopsy. Men with PSAD less than 0.09 ng/ml 2 were unlikely to harbor ISUP grade group 2 or greater disease when PSA was ≤ 20 ng/ml. In patients having PSAD values between 0.09 and 0.19, adjunction of prostate volume increases the performance of PSAD for the detection of clinically significant disease. We propose testing PSAD prior to decision on prostate biopsy since it is a simple inexpensive and available tool that can be used to reduce over detection and morbidity of unnecessary biopsies.

Ethics approval
Soroka University Medical Center ethics committee approved the study and waived informed consent requirements. We confirm that all methods were performed in accordance with the relevant guidelines and regulations.