Introduction

Prostate cancer (PCa) is the most common non-cutaneous malignancy in men and is a major health concern in developed countries. Traditionally, clinical diagnosis is made by digital rectal examination (DRE) in association with serum PSA level and confirmation by 10–12-core systematic biopsies [1]. However, this blind diagnostic method leads on one hand to a risk of under-detection of potentially significant index lesions, on the other hand, systematic biopsies may lead to detection of clinically indolent, low-risk PCa, with a consequent risk of overtreatment [2].

Undeniably, it is known that Gleason score (GS) remains one of the most important prognostic factors and an essential component for determining the best choice of treatment. However, several large studies have shown a limited correlation between systematic biopsies and final histopathologic outcomes, reaching values of 50%, with disease upgrading and downgrading in more than 30% and 20% of cases, respectively [3, 4]. This upgrading can in fact lead to inappropriate or under-treatment in a subgroup of patients [5].

Multiparametric magnetic resonance imaging (mp-MRI) has been studied as a new tool for early detection of prostate cancer before performing initial biopsies and recent studies confirm its high sensitivity especially for clinically significant cancer [6,7,8,9]. Suspicious foci can be targeted during biopsy and several modalities have been developed for clinical practice, including MRI/ultrasound (MRI/US) image fusion and cognitive techniques [10]. Few prospective studies comparing image fusion and cognitive techniques have not clearly shown a significant difference in terms of cancer detection and grade concordance with final specimen [11,12,13]. However, it appears that MRI/US image fusion biopsy shows better reproducibility with a fairly rapid learning curve and can be used to improve accuracy for focal treatments and/or radiotherapy [14,15,16]. Although randomized studies have shown contradictory results in identifying significant difference in terms of cancer detection between targeted and systematic biopsies, it appears that targeted biopsies would improve the accuracy of biopsy [17,18,19,20]. Indeed, it seems that they improve histopathological concordance with the final prostatectomy specimen [21,22,23,24].

The aim of this multicentric study was to evaluate the accuracy in histologic grading of MRI/US image fusion biopsy by comparing histopathology between systematic biopsies (SB), targeted biopsies (TB) and the combination of both (SB + TB) with the final histopathologic outcomes of radical prostatectomy.

Materials and methods

We performed a retrospective multicentric study with a European database (Belgium, Italy, France and UK) on patients who underwent MRI/US image fusion biopsy with Koelis® device (Koelis®, La Tronche, France) between 2010 and 2017. A total of 2115 cases were found, of which 443 underwent radical prostatectomy in the same institution in which biopsies were performed, with complete data available.

Clinical and pathologic data was retrospectively gathered. Mp-MRI was performed prior to biopsy when a prostate cancer was suspected by an elevated PSA and/or a positive DRE although it is not currently recommended by EAU guidelines in biopsy-naïve patients [1]. The mp-MRI protocol was dependent on the institution and followed the European Society of Urogenital Radiology (ESUR) guidelines with at least T2-weighted (T2W), diffusion-weighted imaging (DWI) and dynamic contrast-enhanced (DCE) sequences. Suspicious lesions were defined by a Prostate Imaging Reporting and Data System score (PI-RADS version 1 was used for 110 patients until 2015 which were converted to the PI-RADS version 2 after retrospective analysis of the images, and version 2 thereafter for 333 patients) and then were delineated by the radiologist using the Koelis® software.

Urostation® (93%) and Trinity® (7%) devices (Koelis®, La Tronche, France) were used to create highly detailed 3D maps of the prostate combining elastic MRI/US fusion, 3D ultrasound and Organ-Based Tracking® technology. Biopsies were performed exclusively transrectally by experienced urologists (> 100 biopsies performed) who knew the MRI results before the procedure. All the patients gave informed consent and biopsies were performed according to EAU guidelines; quinolones were used before biopsies and enema was performed according to the operator’s choice [1]. They underwent systematic biopsies (SB) (median 12 cores) and then targeted biopsies (TB) (median 3 cores), depending on the prostate anatomy and size of suspicious lesions. Systematic biopsies were performed using an extended pattern 12-core biopsy. No matter the location or the size of the target area, systematic biopsies were always carried out in the same, standardized manner. Biopsies of hypoechoic area were left to the discretion of operators as well as the transitional zone.

Each prostate biopsy was individualized, labeled and embedded in paraffin. For each case, biopsy and prostatectomy specimens were evaluated by the same uro-pathologist in each institution. For the needle biopsy and prostatectomy specimens, the overall grade was assigned based on the part with the highest Gleason score which is usually associated with the grade of the dominant nodule in concordance with the recommendations of the International Society of Urological Pathology (ISUP) [25]. Gleason scores were reported as five groups (ISUP grades 1–5) as recommended in recent ISUP guidelines. Significant prostate cancer was defined by an ISUP grade equal or superior to 2.

Indication for radical prostatectomy was taken in line with the EAU guidelines [1]. All RPs were performed with pure laparoscopic or robot-assisted laparoscopic approach by experienced urologists, according to the institutional policy.

Cochran’s Q test was used to test for the difference in concordance with final histopathology between systematic (SB), targeted (TB) and combined biopsies (SB + TB). McNemar test was conducted as a post hoc test. Univariable and multivariable logistic regressions were performed to analyze factors predicting histopathological concordance for targeted biopsies. Variables tested included: PSA (categoric < 10 vs. 10–20 vs. > 20), prostate volume (categoric < 30 vs. 30–60 vs. > 60), previous biopsy (categoric), number of targets (continuous), PI-RADS score (categoric 3 vs. 4 vs. 5), diameter of target (continuous), and number of cores per target (continuous). We assumed a significance level of 0.05. Statistical analysis was performed using SPSS version 25 (IBM Corp®, New York, USA).

Results

Patient characteristics are shown in Table 1. In total, 126 cases (28.4%) had positive digital rectal examination and 45% of the patients were biopsy naïve. The median prostate volume was 45.5 cc (IQR 25). Mp-MRI identified a single index lesion with a PI-RADS score of 3, 4 and 5 in 17.6%, 45.6% and 30.9% of the patients, respectively. Of note, there was insufficient data regarding the PI-RADS score for 25 patients (5.6%) because of incomplete protocols at the time of the PI-RADS version 1. The median number of cores taken per lesion was 3 (IQR 2) for targeted biopsy and 12 (IQR 3) for systematic biopsies.

Table 1 Patient characteristics of 443 patients
Full size table

Histopathological concordance with final specimen is shown in Table 2 and Fig. 1a (overall ISUP grade), b (ISUP grade ≥ 2). Cochran’s Q test was performed to test the difference between the three groups: systematic biopsy only (SB group), targeted biopsy only (TB group) and the association of both techniques (SB + TB group). The test was positive in all the categories whether the analysis was for the downgrading (p < 0.001), upgrading (p < 0.001) or concordance (p < 0.001) and it can be concluded that the proportions in at least two of the groups were significantly different from each other. McNemar test was then performed to compare each set of groups head to head. Table 3 illustrates concordance in ISUP grade across biopsy techniques and radical prostatectomy specimen. For overall ISUP grade, a significant difference was found between SB group and SB + TB group for downgrading, upgrading and concordance (p < 0.001). Same results were found between TB group and SB + TB group (p < 0.001). However, there was an absence of significant difference between SB group and TB group (upgrading p = 0.2, downgrading p = 0.2, concordance p = 0.6). For ISUP grade ≥ 2, a significant difference was found between SB group and SB + TB group for upgrading, downgrading and concordance (p < 0.001), between TB group and SB + TB group (upgrading p < 0.001, downgrading p < 0.001, concordance p = 0.001), and between SB group and TB group (upgrading p < 0.001, concordance p = 0.001) except for downgrading (p = 0.2).

Table 2 Histopathologic results
Full size table
Fig. 1
figure 1

Histopathological concordance with final specimen, a overall ISUP grade, b ISUP ≥ 2 and c ISUP ≥ 2 (biopsy-naïve patients)

Full size image
Table 3 Histopathology concordance between biopsy (systematic + targeted) and final specimen
Full size table

A post hoc subgroup analysis including only biopsy-naïve patients with ISUP grade ≥ 2 at radical prostatectomy showed similar results with a significant difference between SB group and SB + TB group (upgrading p < 0.001, downgrading p = 0.03, concordance p < 0.001), between TB group and SB + TB group (upgrading p < 0.001, downgrading p = 0.004, concordance p = 0.02) but there was no significant difference between SB group and TB group (upgrading p = 0.3, downgrading p = 0.5, concordance p = 0.1) (Fig. 1c).

In univariable logistic analysis, previous biopsy [OR = 0.6, 95% CI (0.4–0.9), p = 0.04] was associated to a significant reduction in concordance and number of cores taken [OR = 1.2, 95% CI (1.1–1.3), p = 0.004] was significant predictive factors for concordance (Table 4). Multivariable analysis revealed that these same factors were independently associated with histopathological concordance [previous biopsy OR = 0.6, 95% CI (0.4–0.9), p = 0.01; number of cores taken OR = 1.2, 95% CI (1.1–1.3), p = 0.004].

Table 4 Logistic regression for targeted biopsy concordance
Full size table

Of note, 26 patients (5.9%) had negative prostate biopsies with previous positive biopsies. In this subset, patients were under active surveillance but chose to carry out a surgical intervention after discussion with their surgeon. Final histopathology at radical prostatectomy was 9 ISUP grade 1, 10 ISUP grade 2, 6 ISUP grade 3 and 1 ISUP grade 4. The final TNM stage was 9 T2a, 16 T2c and 1 T3b. In total, nine patients (2%) with grade I (acute urinary retention) and nine patients (2%) with grade II (severe hematuria and prostatitis) complications according to the Clavien–Dindo classification were reported.

Discussion

The last decade has witnessed a revolution in PCa diagnosis, with the introduction of mp-MRI and the inverse stage migration allowing for a better identification of aggressive cancers requiring radical therapy and indolent cancers as potential candidates for active surveillance [6, 26]. The ultimate goal should be to decrease the risk of overtreating young adults and undertreating the older generation [27]. The purpose of mp-MRI is to help the urologist to better characterize suspicious lesions, to be able to delineate potentially clinically significant PCa which can be biopsied later despite having a challenging location such as the apex and the anterior zone of the prostate and eventually reducing the number of useless biopsies for indolent cancer [6, 7]. This technology is even more important because targeted biopsies may be helpful in causing a breakthrough in focal tumor treatment without the need of radical therapy in the absence of definitive histopathology.

Our study found no significant difference in terms of concordance between systematic and targeted biopsies for all the cancers but a better accuracy for targeted biopsy when significant cancers were taken into account. This finding suggests that the use of targeted biopsies alone can detect as accurately as systematic biopsies prostate cancer and support the conclusion of several studies using different platforms of MRI/US fusion (Artemis®, BioJet®, Urostation®) [21,22,23,24]. If we only take into account studies using Koelis® device, Baco found a concordance for targeted biopsy of 70%, an upgrading of 14% and downgrading of 16% in a retrospective analysis of 135 patients [21]. Lanz found similar results with a concordance for targeted biopsy of 67%, an upgrading of 29% and a downgrading of 4% with a retrospective analysis of 125 patients [24]. Of note, these studies evaluated the accuracy of targeted biopsy without a direct comparison with systematic biopsy or the association of both. It is, therefore, reasonable to raise the question of the usefulness of systematic biopsies. Of note, our study revealed that significant prostate cancers were found in cases of negative targeted biopsies with 38 ISUP grade I (8.6%) and 63 ISUP grade ≥ 2 [14.2%: 40 ISUP grade II (9%), 17 ISUP grade III (3.8%), 5 ISUP grade IV (1.1%) and 1 ISUP grade V (0.2%)], a result which appears validated by other studies confirming that targeted biopsies alone might miss 3.8–17% of the significant prostate cancer [28]. Furthermore, around 10% of the tumors were undetected by MRI and in few cases, the delineated target volume does not always reflect the real tumor volume with a 16% over estimation and a 32% underestimation according to Cornud [7, 8, 29]. However, Baco showed a 100% match in terms of localization of targeted biopsies for the index lesion and an overall 95% precision rate with a 5% error margin attributable to missing the tumor on MRI using Koelis® system [21]. Therefore, taking into account systematic biopsies allows adjusting for the lack of precision in tumor volume on MRI and, to a lesser extent, to improve the performance of targeted biopsies. In fact, combining both techniques (systematic and targeted) achieves a 63.2% in concordance with final histopathology, an upgrade of 23.9% and a downgrade of 12.9%. This same trend is seen whenever we take into account the subcategory of significant cancers (ISUP grade ≥ 2) with a significant difference in comparison with the two biopsy techniques separately. This proposal goes in the opposite direction of the conclusion of the PRECISION trial that supports the fact of doing only targeted biopsy [20]. However, the primary outcome of this study was the detection of men with clinically significant cancer and not the histopathologic concordance. Of note, 30 patients underwent radical prostatectomy with sufficient data and percentages of concordance, upgrading and downgrading were similar between targeted biopsy group and systematic biopsy group. To our point of view, the association of both techniques remains the standard for PCa detection and our results reinforce these recommendations [1].

The study included patients over a 7-year period and different versions of the fusion system were used (Urostation® and Trinity® in 93% and 7% of cases, respectively). However, it is known that MRI/US fusion targeted biopsies are associated with a short learning curve [14] and all operators had a vast experience with both targeted and systematic biopsies. Moreover, the histopathologic concordance between targeted biopsies and final specimen for the first 100 cases (2010–2014) and last 100 cases (2016–2017) was 57% and 59%, respectively.

Although recommendations have been proposed at the International Society of Urological Pathology (ISUP) consensus meeting in 2005 and 2014, discordance among pathologists remains in assigning Gleason score and ISUP grades [25]. In the case of multiple cores having different Gleason grades, some pathologists choose to give an overall Gleason score rather than the highest Gleason score. In general, the pathologist should report the grades of core separately and in our study, we used the highest Gleason score (biopsy and final specimen) to calculate the histopathologic concordance. Nonetheless, as the same pathologist performed the analysis for the biopsy and the final prostatectomy specimen, this discordance is limited.

The number of cores taken during targeted biopsy was an independent predictive factor for concordance with final histopathology and it is currently recognized that two cores per lesion are generally sufficient while the median number was 3 per lesion in our study. Number of cores depends on lesion sizes and appreciation of the operator and it is important to note that multiple studies confirmed the absence of correlation between the number of cores taken and the potential risk of infection or bleeding [30]. Given our results, we advise urologists should not hesitate to perform additional biopsies to maximize the accuracy of diagnosis.

Although this is to our knowledge the largest study comparing histopathology between biopsy and final specimens, we acknowledge its limitations. This is a retrospective study with heterogeneity in the data collection and looks at a population who underwent radical prostatectomy with a significant disease. Mp-MRI and biopsies were performed by a large number of specialists with a variability of experience and habits in different centers without a central pathologic revision. Urologists were also not blinded to MRI results so there could be an influence on how systematic biopsies were performed. However, the same pathologist for each center evaluated the biopsies and the correlated prostatectomy specimen, thus such bias is reduced.

Conclusions

In this retrospective study, MRI/US image fusion and systematic biopsies were found to be complementary, significantly increasing concordance with final histopathology and causing a significant decrease in disease upgrading. Combining these two techniques may aid in tailoring the adequate treatment for each patient. Prospective studies are awaited to validate our findings.