Introduction

Treatment decision-making in prostate cancer (PCa) is mainly based on accurate, clinically meaningful, reproducible risk classifications including prognostic parameters such as PSA, clinical stage, and grade group on biopsies [1,2,3]. The EAU risk classification that classified newly diagnosed PCa within three well-established categories (low, intermediate, and high risk) aims at predicting the risk of recurrence after curative treatment and adapt PCa management according to that risk [1,2,3]. However, the EAU risk classification has been validated in cohorts of patients undergoing systematic, not imaging-targeted biopsies (SB). Recent literature has demonstrated the superiority of targeted biopsies (TB) over SB for the detection of clinically significant PCa [4,5,6,7]. This new diagnostic pathway modifies grade group distribution with more high-grade disease detected [8]. This could lead to important changes in the pre-treatment prognostic models with less marked survival differences between EAU risk groups.

We acknowledge that the widely used pre-treatment risk classification may no longer apply to men undergoing MRI-guided biopsy. Therefore, given the increasing utilization of MRI-guided biopsy in contemporary practice, we think it is critical to revise prognostic tools to account for MRI-targeted biopsy [9]. Few studies have assessed the improvement of prediction of adverse pathology prediction at the time of radical prostatectomy by MRI and TB [10]. However, to our knowledge, no series of patient undergoing TB has included survival outcomes after surgery and has validated recurrence risk prediction models.

Therefore, we aimed to evaluate the performance of the EAU risk classification in a cohort of patients undergoing SB and fusion TB and to compare it with its accuracy in a contemporary cohort of patients who were diagnosed by a standard, not imaging-guided biopsy pathway.

Materials and methods

Study population

Between January 2014 and May 2019, 526 patients underwent an RP for pathologically biopsy-proven prostate cancer after a pre-biopsy positive (PI-RADS ≥ 3) multiparametric (mp) MRI (“targeted biopsy” or TB cohort). During the same period, 819 patients underwent RP after a diagnosis based only on systematic biopsy (“standard biopsy” or SB cohort), because of pre-biopsy negative MRI or absence of imaging prior to biopsy. For these patients, MRI could have been performed after the diagnosis for staging purposes and did not influence the biopsy procedure.

In the TB cohort, the imaging protocol consisted of multiplanar T2-weighted images, diffusion-weighted imaging, dynamic contrast-enhanced MRI, and T1-weighted images with fat suppression according to the European Society of Urogenital Radiology guidelines [11]. Both institutions used a 1.5-T MR unit and a 16-channel coil. No endorectal coil was used. The maximal b value used for diffusion-weighted imaging was b 2000. The mpMRI images were scored and reported according to Prostate Imaging-Reporting and Data System v.2 (PI-RADS) [12]. MRI lesions with a PI-RADS ≥ 3 on mpMRI were submitted to targeted biopsy using real-time transrectal ultrasound (TRUS) guidance via a software registration system with elastic fusion (Koelis® system). A minimum of two targeted cores were taken for each suspicious lesion on mpMRI. All patients also underwent concomitant systematic biopsy at the time of the targeted biopsy, with at least ten random cores taken.

For all patients from both cohorts, RP was performed by high-volume surgeons in one of the two study centres. Delay between biopsy and RP was < 6 months (median 3.2 months, from 0.8 to 5.7). All imaging and biopsy procedures have been performed in two institutions by radiology and urology seniors experienced in prostate cancer diagnosis. Biopsy and RP specimens were evaluated by senior dedicated uropathologists from the two institutions. PSA recurrence was defined by any PSA above 0.2 ng/ml. Biochemical follow-up was standardized with a PSA test at 6 weeks, 3 months and then every 6 months after surgery. No patient received any adjuvant treatment. Data from clinical evaluation, biopsy and RP specimens, and follow-up were recorded in a prospective database.

Analyses

The clinical, biological, and pathological findings were assessed in the overall population, and compared between the two cohorts (SB versus TB cohorts).

The primary endpoint was the prognostic performance of three-group EAU risk classification for predicting biochemical recurrence-free survival (RFS) in both cohorts. A multivariable Cox regression analysis was performed to develop a new risk classification which could outperform the EAU classification in the context of TB diagnostic pathway.

Statistics

The qualitative data were tested using a Chi-square test, or Fisher’s exact test as appropriate and the continuous data were tested using Student’s t test. The Mann–Whitney’s test was used in case of no normal distribution. Survival curves were assessed using Kaplan–Meier method and multivariable analyses were run using Cox regression models. ROCs were also built to assess AUC values. The limit of statistical significance was defined as p < 0.05. The SPSS 22.0 (Chicago, Illinois) software was used for analysis.

Results

Overall population characteristics

Main patient characteristics are listed in Table 1. We found significant variability in the number of patients within each cancer risk category due to TB when compared to SB (p < 0.001). Overall, 61% of low-risk PCa cases defined by SB alone were reclassified as intermediate-risk cases by adding MRI-TB. PCa cases classified as intermediate risk using only SB were reclassified as high risk in 7% of cases when adding MRI-TB.

Table 1 Baseline characteristics of the cohort and comparisons of pathological/follow-up parameters between standard versus targeted biopsy cohorts
Full size table

Biopsy features’ comparison between SB and TB cohorts is also shown in Table 1. The mean number of biopsy cores taken was superior in the SB cohort (16.2 versus 14.7, p < 0.001) compared with the TB cohort, but the mean number of positive cores was lower (5.4 versus 5.9, p = 0.016). Grade group 2 and 3 was reported in 52.1% and 22.6% of cases, respectively, in the TB cohort, compared with 39.1% and 13.9%, respectively, in the SB cohort (p < 0.001).

Pathological outcomes

Pathological features in RP specimens are reported in Table 2, as well as the comparisons between both cohorts. Main pathological features were comparable in the two cohorts. Grade group on final pathology did not differ significantly between the SB and TB cohorts (p = 0.751), as well as pT (p = 0.501) and pN stage (p = 0.101). Patients in the TB cohort had a lower risk of upgrading (31.0% versus 48.2%, p < 0.001) with an improved concordance rate (53.2% versus 43.5%) compared with those included in the SB cohort. The proportion of high-grade disease within each risk group significantly differed according to the biopsy pathway. The rate of grade group ≥ 3 or ≥ 4 disease was significantly lower in the TB cohort. In the SB cohort, 22.2% of grade group ≥ 3 disease was reported in the low-risk group, as compared with only 8.9% in the TB cohort low-risk group (p < 0.001).

Table 2 Rates of unfavourable pathological outcomes according to the risk group classification in the two cohorts
Full size table

Recurrence-free survival outcomes

Mean follow-up after RP was 26.6 months, and comparable between the two cohorts (25.8 versus 27.0 months, p = 0.125). RFS curves did not differ significantly between the SB and the TB cohorts (Fig. 1a, log rank test; p = 0.538). The 2-year RFS rates were 84.9% and 88.1% in the SB and TB cohorts, respectively.

Fig. 1
figure 1

a Biochemical recurrence-free survival curves (bRFS) stratified by the biopsy pathway: standard (SB) versus targeted biopsy (TB) cohort (p = 0.538). b Biochemical recurrence-free survival curves (bRFS) stratified by the risk group classification and by the cohort. LR low risk, IR intermediate risk, HR high risk

Full size image

Figure 1b shows the RFS curves stratified by two factors, the EAU risk classification and the cohort. In the SB cohort, the EAU risk classification was also significantly predictive for PSA outcomes in the overall cohort (p < 0.001). The risk of biochemical recurrence was increased by fourfold in intermediate (95% CI 1.6–10) and by 15.7-fold (95% CI 6.2–39.8) in high-risk groups, compared with low-risk group.

In the TB cohort, the EAU risk classification was also significantly predictive for PSA outcomes in the overall cohort (p < 0.001). However, only the high-risk group curve was significantly different with a HR 6.5 (95% CI 1.5–27.5, p = 0.011). Indeed, the risk of biochemical recurrence did not differ significantly between the low- and the intermediate-risk group (p = 0.791; HR 1.2; 95% CI 0.28–5.2) in the TB cohort.

PSA, clinical stage and grade group on biopsy were included into a multivariable Cox regression model in the TB cohort (Table 3). The aim was to develop a new risk classification to improve the accuracy of the prognostic model within the TB cohort. The new risk groups were defined as follows: PSA < 10 and grade group 1–2 and T1–T2a clinical stage (low risk); PSA 10–20 or grade group 3 or T2b clinical stage (intermediate risk); PSA ≥ 20 or ≥ T2c clinical stage or grade group 4–5 (high risk).

Table 3 Multivariable Cox regression model assessing the predictive factors for biochemical recurrence in the TB cohorts
Full size table

The new risk classification outperformed the EAU risk classification in the TB cohort for predicting biochemical recurrence (Fig. 2a). The risk of biochemical recurrence was increased by fourfold in intermediate (p = 0.009) and by 15-fold (p < 0.001) in high-risk groups, compared with low-risk group (Table 3).

Fig. 2
figure 2

a Biochemical recurrence-free survival curves (bRFS) in the TB cohort, stratified by the new classification. LR low risk: PSA < 10 and grade group 1–2 and T1–T2a clinical stage; IR intermediate risk PSA 10–20 or grade group 3 or T2b clinical stage; HR high risk: PSA ≥ 20 or T3 clinical stage or grade group 4–5. b Biochemical recurrence-free survival curves (bRFS) in the TB cohort, stratified by the new 4-group classification. LR low risk: PSA < 10 and grade group 1–2 and T1–T2a clinical stage; IR intermediate risk: PSA 10–20 or grade group 3 or T2b clinical stage; HR high risk: PSA ≥ 20 or grade group 4; VHR very high risk: ≥ T2c clinical stage or or grade group 5

Full size image

Then, we developed a new four-group classification to better define patients at very high risk of recurrence (Fig. 2b, log rank test: p < 0.001). The four risk groups were defined as follows: PSA < 10 and grade group 1–2 and T1–T2a clinical stage (low risk); PSA 10–20 or grade group 3 or T2b clinical stage (intermediate risk); PSA ≥ 20 or grade group 4 (high risk); ≥ T2c clinical stage or grade group 5 (very high risk). The risk of biochemical recurrence was significantly increased by fourfold in intermediate, by ninefold in high-risk groups, and by 29-fold in very high-risk groups, compared with low-risk group (p < 0.001, Table 3). The overall 2-year recurrence rates significantly differed between new classification groups: 1.9%, 8.7%, 16.9%, and 42.9% in the low-, intermediate-, high-, and very high-risk groups, respectively. AUC for predicting recurrence was calculated for each classification. The AUC of the EAU risk classification was 0.686 compared with 0.734 for the AUC of the new risk classification. The improvement of AUC obtained by the new risk classification compared with the EAU classification was strictly comparable in both centres: + 0.048 in centers 1 and 2.

Discussion

MRI followed by TB directed to suspicious lesions has also been shown to optimize PCa prognostic assessment and risk stratification compared with SB alone [8, 13,14,15,16,17,18,19]. However, despite the growing body of knowledge highlighting the benefits of performing MRI and MRI-TB to SB, understanding is still lacking on the effects of their implementation into patient care and prognostic assessment of newly diagnosed disease. Thus, the pathway including MRI followed by TB improves risk stratification of patients suspected of having PCa and has the potential to alter clinical decision-making for PCa patients. However, to date, treatment decision-making is mainly based on risk classifications that have been defined and validated in cohorts of patients undergoing systematic, not imaging-targeted biopsies [1,2,3]. Current tools predicting recurrence after curative treatment and guiding PCa management are driven by the results of systematic, not imaging-guided biopsies. We acknowledge the fact that these models do not apply for targeted biopsies which improve the grade group assessment and thereby, modify risk stratification obtained by SB only. Thus, the accuracy of well-established risk classification should be re-assessed in cohorts of patients undergoing both SB and TB, or at least TB.

Our study corroborated current body of the literature in showing that the use of TB decreased the risk of upgrading, and improved the grading concordance between biopsy and RP specimens [20]. Patients in the TB cohort had a lower risk of upgrading (31.0% versus 48.2%) with an improved concordance rate (53.2% versus 43.5%) compared with those included in the SB cohort. This supported the improvement of initial assessment of suspected PCa in terms of diagnosis, staging, and appropriate PCa management. This also led to a significant variability in the number of patients within each cancer risk category due to TB when compared to SB (p < 0.001). Our findings were in line with previous published reports [21, 22].

Interestingly, although the SB and the TB were strictly comparable in terms of final pathology and PSA outcomes, prognosis of each risk category was statistically different according to the biopsy pathway. As expected, the standard risk classification accurately predicted recurrence outcomes in the SB groups, with three well-separated groups in terms of survival curves. However, such a stratification was not relevant in the TB cohort, with statistical similarity between low- and intermediate-risk groups (p = 0.791). This could be explained by a not negligible proportion of grade group 1 on SB that were reclassified as grade group 2 disease by MRI-targeted biopsies (61%).

Patients in the TB cohort had a lower risk of upgrading (31.0% versus 48.2%, p < 0.001) with an improved concordance rate (53.2% versus 43.5%) compared with those included in the SB cohort.

Widely used pre-treatment risk classification may no longer apply to men undergoing MRI-guided biopsy. And given the increasing utilization of MRI-guided biopsy in contemporary practice, we thought it was critical to revise prognostic tools to account for MRI-targeted biopsy.

Given this contamination between low- and intermediate-risk groups in the TB cohorts leading to modified prognosis as compared with that calculated in the SB cohort, we built new risk classification by pooling grade groups 1 and 2 cases into the same risk group. This led to three new well-separated survival estimates. We also developed a four-group classification to better identify patients at very high risk of early recurrence who could benefit from more aggressive adjuvant strategies.

Improved prognostic accuracy by TP may help for a better counseling of PCa patients during treatment decision-making and for anticipating the potential need for multimodal strategies in very high-risk disease. First, the added value of TB for risk stratification may also play a role on the decision to perform lymphadenectomy and on the guidance of its extent. Gandaglia et al. have recently updated the Briganti’s nomogram which defined the risk of lymph node involvement and the need for lymph node dissection [23]. The integration of MRI and TB into a new nomogram significantly improved its accuracy, in addition to other standard parameters. All patients in the present study had a positive MRI and underwent targeted biopsies. Therefore, we could not guarantee that this new risk classification was also accurate for patients having a negative MRI. It would be interesting to assess the performance of standard/new risk classifications in the subset of patients with negative MRI. Unfortunately, our dataset only included MRI-positive patients which represent about 80% of contemporary patients receiving biopsies according to recent prospective trials [4, 7].

Second, the improvement of risk stratification may modify treatment planning before radiotherapy decision-making. Reed et al. reported that TB led to intensified strategies in 16% of patients (Reed). Dix et al. demonstrated that TB resulted in 13% more androgen deprivation therapy regimens and in a 26% increase of more aggressive radiotherapy treatments compared to SB pathology results [22]. Nevertheless, no final pathology based on RP specimens was available in such series.

The main limitation was the lack of strong clinical endpoints, such as metastasis-free, cancer-specific or overall survival. However, the recommendations favouring the MRI pathway have been recently updated and we cannot get enough follow-up to assess such endpoints. Moreover, EAU guidelines support the use of such recurrence risk prediction tools for treatment decision-making [1]. The two-center design of our study may be a limitation. Nevertheless, all radiologists and biopsy operators involved in the study were highly experienced and beyond their learning curves since the beginning of the study period. The same fusion computer-assisted software was used in the two institutions which reduced interpretation biases [24]. Pathology was not formally re-reviewed, but was interpreted by senior, fellowship trained, uropathologists in both centers. Analyses were not altered after stratification by center status.

Conclusions

MRI-TB mainly leads to reclassification to a higher risk group, which is likely to be accompanied by important changes in risk group distribution and pre-treatment prognostic assessment compared with a diagnostic pathway only based on SB. Thus, the EAU risk group classification which is historically based on tumor grade obtained by SB does not accurately dichotomize low- and intermediate-risk PCa in terms of biochemical recurrence, when MRI and TB are integrated into the diagnostic pathway. The new easy-to-use three-group classification we propose includes grade group 2 cancer within the low-risk group and improves the risk prediction of biochemical recurrence for patients undergoing a TB-based diagnostic pathway. We have also built a four-group classification that better defines very high-risk patients (grade group 5 or clinical ≥ T2c disease) who could benefit from more aggressive adjuvant therapies. External studies and longer follow-up are needed to confirm the superiority of this new classification for recurrence prediction in the opening era of MRI and TB-driven biopsy pathway.