Comparative Analysis of PSA Density and an MRI-Based Predictive Model to Improve the Selection of Candidates for Prostate Biopsy

Simple Summary Magnetic resonance imaging (MRI)-associated prostate-specific antigen density (mPSAD) and MRI predictive models have been proposed for improving the selection of candidates for prostate biopsy among men with suspected prostate cancer (PCa). While the calculation of mPSAD only requires a simple division, the individual risk assessment of PCa using the available risk calculators is also a swift process. We aim to compare the clinical usefulness of mPSAD and an MRI predictive model that utilises the same predictors as the recently developed and externally validated Barcelona MRI predictive model (MRI-PMbdex). Abstract This study is a head-to-head comparison between mPSAD and MRI-PMbdex. The MRI-PMbdex was created from 2432 men with suspected PCa; this cohort comprised the development and external validation cohorts of the Barcelona MRI predictive model. Pre-biopsy 3-Tesla multiparametric MRI (mpMRI) and 2 to 4-core transrectal ultrasound (TRUS)-guided biopsies for suspicious lesions and/or 12-core TRUS systematic biopsies were scheduled. Clinically significant PCa (csPCa), defined as Gleason-based Grade Group 2 or higher, was detected in 934 men (38.4%). The area under the curve was 0.893 (95% confidence interval [CI]: 0.880–0.906) for MRI-PMbdex and 0.764 (95% CI: 0.774–0.783) for mPSAD, with p < 0.001. MRI-PMbdex showed net benefit over biopsy in all men when the probability of csPCa was greater than 2%, while mPSAD did the same when the probability of csPCa was greater than 18%. Thresholds of 13.5% for MRI-PMbdex and 0.628 ng/mL2 for mPSAD had 95% sensitivity for csPCa and presented 51.1% specificity for MRI-PMbdex and 19.6% specificity for mPSAD, with p < 0.001. MRI-PMbdex exhibited net benefit over mPSAD in men with prostate imaging report and data system (PI-RADS) <4, while neither exhibited any benefit in men with PI-RADS 5. Hence, we can conclude that MRI-PMbdex is more accurate than mPSAD for the proper selection of candidates for prostate biopsy among men with suspected PCa, with the exception of men with a PI-RAD S 5 score, for whom neither tool exhibited clinical guidance to determine the need for biopsy.

was created from the development cohort of 1486 men studied at Vall d'Hebron Hospital (VHH) and the external validation cohort of 946 men from Parc de Salut Mar (PSM) as well as Hospital Germans Trias i Pujol (GTiPH). This study was conducted between January 1, 2006, and 31 December 2019 [38]. Pre-biopsy 3-Tesla mpMRI and guided or systematic prostate biopsies were always performed. Men undergoing 5-α reductase inhibitor treatment due to symptomatic benign prostatic hyperplasia, having a previous diagnosis of PCa, exhibiting isolated atypical small acinar proliferation, and exhibiting high-grade prostatic intraepithelial neoplasia with atypia were excluded. Furthermore, written consent for prostate biopsy was obtained from all participants, and the project was approved by the institutional review board of VHH (PRAG-317/2017).

Intervention
The development of MRI-PMbdex and the individual generation of csPCa likelihood was expressed as percentages ranging from 0 to 100%. mPSAD (ng/mL 2 ) was calculated from the pre-biopsy serum PSA and the prostate volume reported in the pre-biopsy MRI, and individual generation of csPCa likelihood was expressed as percentages ranging from 0 to 100%.

MRI Technique and Evaluation
Magnetic resonance scans were acquired on 3-Tesla scanners with a standard surface phased-array coil. Magnetom Trio (Siemens Corp., Erlangen, Germany) equipment was used in VHH, Diamond Select Achieva (Phillips Corp., Eindoven, The Nederland) in PSM, and Nova Dual (Phillips Corp., Eindoven, Nederland) in GTiPH. The acquisition protocol included T2-weighted imaging (T2W), diffusion-weighted imaging (DWI), and dynamic contrast-enhanced (DCE) imaging, in accordance with the European Society of Urogenital Radiology guidelines [54]. In each institution, an expert radiologist, with over five years of experience and over 300 mpMRI reported per year, analysed the images and reported them according to PI-RADS v2.0, using a 5-point likelihood scale for csPCa [55]. Complex cases were reviewed by two expert radiologists.

Prostate Biopsy Procedure
All men from the participant institutions underwent 2 to 4-core mpMRI-TRUS cognitive fusion-guided biopsies of suspicious lesions and 12-core TRUS systematic biopsies when the PI-RADS reported in the pre-biopsy mpMRI was 3 or higher, while 12-core TRUS systematic biopsies were performed when the PI-RADS was lower than 3 [56]. In each institution, the biopsies were performed by an experienced urologist, with over five years of experience and who had performed over 300 biopsies per year, using a BK Focus 400 ultrasound scanner (BK Medical Inc. Herlev, Denmark) in VHH, Siemens Acuson 150 (Siemens Inc., Erlangen, Germany) in PSM, and a Sonolite Antares (Siemens Inc., Erlangen, Germany) in GTiPH.

Pathologic Analysis and csPCa Definition
The biopsy samples were sent separately to each pathology department, where an expert uro-pathologist analysed the biopsy specimens and, after identifying the PCa, reported the International Society of Uro-Pathology (ISUP) Gleason-based Grade Group (GG) [57]. Complex cases were analysed by two expert uro-pathologists. Any ISUP-GGG ≥ 2 was defined as csPCa.

Endpoint Measurements
The endpoint measurements were the rates of detected csPCa and missed csPCa, in addition to the frequency of avoided prostate biopsies and iPCa.

Statistical Analysis
The reporting recommendations for tumour marker prognostic studies (REMARK) [58] and the update of standards for reporting diagnostic accuracy studies (STARD 2015) [59] were followed. Medians and interquartile ranges (25-75 percentiles) were used to describe the quantitative variables, and rates were used to describe the qualitative variables. The MRI-PMbdex individual likelihoods of csPCa were generated from the logistical regression analysis performed from the included independent predictors of the Barcelona MRI-PM [38]. The mPSAD individual likelihoods of csPCa were generated from the logistic regression analysis performed from the following independent predictors: PI-RADS score and mP-SAD. The chi-square and Mann-Whitney tests were used to find associations between the variables. Receiver operating characteristic curves (ROC) and areas under the curve (AUC) were used to analyse the efficacy of mPSAD and MRI-PMbdex for csPCa detection; the DeLong test was conducted to compare the AUCs [60,61]. Furthermore, decision curve analysis (DCA) was used to discern the net benefits of mPSAD and MRI-PMbdex over performing biopsies in all men [62]. The thresholds of mPSAD and MRI-PMbdex were selected from the 95% sensitivity for csPCa, and the specificities were analysed and compared. The performances were analysed based on sensitivity, specificity, negative and positive predictive values, accuracy, avoidable biopsies, and potentially missed csPCa. In addition, odds ratios and 95% confidence intervals (CI) were assessed. Tests with two-sided p < 0.05 were considered statistically significant. The statistical analyses were computed using the R programming language v.4.0.3 (The R Foundation for Statistical Computing, Vienna, Austria), and SPSS v.25 (IBM, Statistical Package for Social Sciences, San Francisco, CA, USA) was used.

Characteristics of the Study Population
The characteristics of the entire population are summarised in Table 1. A median age of 68 years, serum PSA of 6.5 ng/mL, and prostate volume of 55 cm3 can be observed. The DRE was abnormal in 25% of the men; 28% were scheduled to repeat prostate biopsies, and 6.6% had PCa family history. PCa was diagnosed in 1214 men (49.9%), csPCa in 934 men (38.4%), and iPCa in 280 men (11.5%).
The characteristics of the men according to PI-RADS categories as well as comparisons between each pair of consecutive PI-RADS are presented in Table 2. All characteristics showed differences regarding all PI-RADS categories (p < 0.001), with the exception of PCa family history, which showed a non-significant trend that ranged from 4.7% in men with PI-RADS 1 to 8.3% in men with PI-RADS 5 (p = 0.287). On comparing the characteristics of men with PI-RADS 1 and those with PI-RADS 2, it was found that only the frequency of repeat biopsies was significantly higher among the latter (19.4% vs. 34.1%; p < 0.001); all other characteristics were similar between the two categories. Between PI-RADS 2 and PI-RADS 3, the frequency of PCa detection increased from 20.3% to 30.5% (p = 0.028), while the frequency of csPCa and iPCa exhibited no significant increase. Men with PI-RADS 4 were significantly older than those with PI-RADS 3, had a higher PSA and mPSAD, had a lower prostate volume, and exhibited an abnormal DRE more often. The frequency of repeat biopsies, PCa family history, and iPCa detection was similar in both PI-RADS 3 and PI-RADS 4 categories; however, a significant increase in csPCa frequency was observed (p < 0.001). Moreover, in PI-RADS 5 compared with PI-RADS 4, the frequency of repeat biopsies and iPCa was low, while the frequency of csPCa and overall PCa increased significantly (p < 0.001). The frequency of repeat biopsies as well as iPCa also decreased, while the frequency of all PCa and csPCa increased significantly (p < 0.001).

Performance of mPSAD and MRI-PMbdex in the Entire Population
The efficacy of MRI-PMbdex in detecting csPCa was higher than that of mPSAD. The AUC was 0.893 (95% CI: 0.880-0.906) for MRI-PMbdex and 0.763 (95% CI: 0.774-0.783) for mPSAD, with p < 0.001 ( Figure 1A). The MRI-PMbdex exhibited net benefit over mPSAD and over biopsy of all men from a csPCa probability threshold of 2%, while mPSAD exhibited net benefit over biopsy of all men from a csPCa probability threshold of 18% ( Figure 1B).

Discussion
First, the characteristics of suspected PCa regarding PI-RADS categories are considered. The similar characteristics of men with PI-RADS 1 and PI-RADS 2, in addition to similar csPCa detection rates (9.8% vs. 10.6%) and high iPCa over detection rates (above 50%) in both subsets, justify considering these two PI-RADS categories as a negative mpMRI. The observed 90% negative predictive value of mpMRI was in the recently reported ranges (i.e., 80-95%) [10][11][12]. The characteristics of suspected PCa with PI-RADS 3 were closer to those with PI-RADS 2 than those with PI-RADS 4. Moreover, the csPCa detection rate with PI-RADS 3 (16.9%) was closer to that of men with PI-RADS 2 (10.8%) than that of men with PI-RADS 4 (52.2%). Therefore, PI-RADS 3 is an uncertain scenario more similar to a negative MRI than to PI-RADS 4; this justifies why some authors consider men with PI-RADS < 4 as candidates to avoid prostate biopsy [9,15]. Finally, the characteristics of those with PI-RADS 5 were notably different from those with PI-RADS 4. Furthermore, the csPCa risk increased with PI-RAD 5, reaching 83.6%. Taking these factors into account, it is clear that considering PI-RADS > 3 in the same group as the other categories is inadvisable. Hence, the data of men with PI-RADS 4 and PI-RADS 5 should be considered separately [5].
PSAD was introduced by Benson et al. in 1992 to improve the specificity of serum PSA in distinguishing men with localised PCa from those with benign prostatic enlargement [17,63]. Shortly after, it was incorporated by Catalona et al. as a tool for prostate biopsy decision-making in the early detection of PCa [64]. The inverse relationship between prostate volume and PCa risk has been recently confirmed in a systematic review of the literature from the last thirty years [65]. Transrectal ultrasonography was recommended for assessing the prostate volumetry for PSAD calculation from the beginning, as it yields a more accurate assessment than suprapubic ultrasonography [66]. Most studies analysing the clinical usefulness of PSAD have assessed the prostate volume just before TRUS systematic biopsies [67][68][69][70]. However, routine TRUS prostate volume assessment for calculating PSAD is not typically performed for prostate biopsy decision-making. The spread ERSPC (European Randomised Screening Prostate Cancer) risk calculator has incorporated the highly subjective prostate volume estimation from DRE [71,72]. The main reason for the current resurgence of PSAD may be the spread of pre-biopsy MRI [73], which provides the most accurate prostate volume assessment without additional cost; this helps avoid TRUS, which is time-consuming and cumbersome [74]. mPSAD has been directly incorporated as ng/mL 2 into some MRI-PM [23,26,32,33,75] or indirectly incorporated from serum PSA and prostate volume into other MRI-PM [28,38,[76][77][78][79]. Logistical regressions performed to develop MRI-PM have shown PSAD to be the most powerful predictor of csPCa after the PI-RADS score. Recent studies analysing the clinical usefulness of mPSAD for csPCa detection according to PI-RADS categories have shown its dynamic behaviour, which suggests the need for different thresholds to obtain similar predictive values in different PI-RADS categories [21,80]. Nevertheless, the proper mPSAD thresholds for specific PI-RADS categories should be selected from each area, given the specific characteristics of cared populations of men with suspected PCa [22,81]. The most recent MRI-PMs share the latest PI-RADS versions, with direct or indirect mPSAD, and may share other independent clinical predictors, such as age, PCa family history, initial or repeat biopsy, and DRE.
Our study is the first to compare mPSAD and MRI-PM in a head-to-head manner for improving the selection of candidates for prostate biopsy in suspected PCa men after mpMRI. The compared MRI-PMbdex shared the same predictors of csPCa as the recent Barcelona MRI-PM. MRI-PMbdex and mPSAD avoid the inconveniences of modern markers as they are free of cost, do not require sample procurement, and require very little time for the assessment. This study was carried out on a sizable, multicentre population representing a metropolitan area. In all participant institutions, the criteria for PCa suspicion and diagnostic approach adhered to the EAU PCa guidelines [7]. MRI-PMbdex outperformed mPSAD in the entire population, and it exhibited net benefit over mPSAD, which, in turn, exhibited a slight benefit over performing biopsies on all men. The behaviour of both tools regarding PI-RADS categories showed that MRI-PMbdex outperformed mPSAD in each PI-RADS ≤ 4 category, with different degrees of benefit. However, clinical benefit was found for any tool in men with PI-RADS 5. Additionally, mPSAD did not detect more aggressive tumours compared to MRI-PMbdex in men with PI-RADS ≥ 4. The different behaviour of mPSAD [21] and MRI-PMs [38] regarding PI-RADS categories is the consequence of the incidence of csPCa among them. We selected different thresholds of mPSAD to obtain similar performance in our series of PI-RADS 3 than that of Görtz et al. [22,81]. The use of appropriate thresholds in different populations has been previously described for mPSAD [19][20][21][82][83][84][85][86][87][88][89][90][91][92][93][94][95].
The limitations of this study include the design, which is prone to sampling and selection biases. Furthermore, no concordance analysis was performed by the pathologists, radiologists, and urologists involved in the study. In addition, the inter-variability that is inevitable among researchers involved in multicentric studies also exists. Moreover, prostate biopsies and mPSAD suffer from intrinsic limitations related to the heterogeneity of prostate cancer and prostate volume. The lack of histopathological analysis of surgical specimens is another limitation, although not every PCa patient undergoes radical prostatectomy. While the definition we adopted for csPCa is the most common one, it may not represent the true csPCa. Moreover, as the MRI-PMbdex is not the developed Barcelona MRI-PM, the results cannot be attributed to this model. Hence, further studies analysing the clinical usefulness of mPSAD and MRI-PMs according to PI-RADS categories are needed.
It will be important in the future to include the results of radiomics and radio-genomics to improve the next generation of MRI-PMs. It appears likely that we will have more powerful predictors than PI-RADS and a more refined definition of csPCa based on genomic expression in addition to the current the pathology of prostate biopsies or surgical specimens [96,97].

Conclusions
MRI-PMbdex, which used the same predictors as the recent Barcelona MRI-PM, outperformed mPSAD in the proper selection of candidates for prostate biopsy. MRI-PMbdex exhibited net benefit over mPSAD in men with PI-RADS ≤ 4. However, regarding men with PI-RADS 5, neither tool showed clinical benefit over biopsy. The clinical usefulness of tools that improve the selection of candidates for prostate biopsy in the entire population does not represent those observed in each PI-RADS category.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.