Abstract
Background/Aim: The purpose of this study was to examine the prognostic value of Prostate imaging-reporting and data system (PI-RADS) v2.1 scoring system in patients who underwent radical prostatectomy (RP). Patients and Methods: Clinical data of 294 patients who received RP between 2006 and 2018 were reviewed and multiple parameters including PI-RADS v2.1 score were employed to identify predictive factors for biochemical recurrence (BCR). Tumor volume was calculated from prostatectomy specimens. Results: Median age at operation and initial PSA level were 67 years old and 7.68 ng/ml, respectively. 44.9 and 24.8% of patients were diagnosed with PI-RADS score 4 and 5 prior to biopsies, respectively. BCR was observed in 17% of patients and median observation period was 63.43 months. After multivariate analysis, PI-RADS v2.1 score 5 [hazard ratio (HR)=2.24, p=0.0124] was an independent predictive factor of BCR in addition to clinical T stage (≥2c) (HR=2.32, p=0.0093) and biopsy Gleason score (≥8) (HR=2.81, p=0.0007). Furthermore, PI-RADS score 5 significantly stratified the prognosis in D’Amico intermediate- and high-risk groups (p=0.0174 and p=0.0013, respectively). We established novel risk classifications including PI-RADS v2.1 score and found that prognostic capabilities were improved as compared to the D’Amico classification. Conclusion: The PI-RADS v2.1 score exhibited significant prognostic value in patients with localized prostate cancer following RP. Risk classifications based on PI-RADS v2.1 score might provide better ability for predicting oncological outcomes as compared to the D’Amico classification system.
Prostate cancer is the most commonly diagnosed cancer among men with approximately 280,000 cases and the second leading cause of cancer death with 34,700 cases in men in the United States (1, 2). Clinically localized PCa has an extremely favorable prognosis, however, early biochemical recurrence (BCR) may occur in some cases after definitive treatment, requiring adjuvant therapy. Previous studies have attempted to develop useful models to predict prognosis after radical prostatectomy (RP). For instance, the D’Amico classification system was proposed as an optimal staging system following local treatment such as RP and radiotherapy (3). Thereafter, this classification including Gleason score sum (GS), prostate-specific antigen (PSA) level, and clinical T stage, has been widely used for risk stratification in patients with localized PCa. However, better risk-based classification systems using preoperative factors are required to enable better informed decision-making regarding treatment.
The Prostate Imaging-Reporting and Data System (PI-RADS) scoring system was first proposed to represent cancer lesion and aggression using muti-parametric magnetic resonance imaging (mp-MRI) in 2012 by the European Society of Urogenital Radiology (ESUR) (4). Thereafter, several improvements have been made, and version 2.1 is currently in clinical use (5, 6). This system assesses the detectable lesions by mp-MRI and scores the degree of clinically significant cancer lesions (4). The PI-RADS v2.1 scoring system has previously been reported to detect clinically significant prostate cancer, improve the diagnostic accuracy, and avoid unnecessary prostate biopsies (7). However, little is known about its clinical significance as a prognostic predictor after definitive treatment. Given that the PI-RADS scoring system is relevant to likelihood of prostate cancer, in other words, might be related to the size of the cancer lesions on MRI findings, it may be useful to develop a risk classification prior to local therapy for predicting oncological outcomes. The D’Amico classification system was originally developed in 1998 and has since been widely applied in approaches for assessing prostate cancer risk (8-10). This classification system was designed to evaluate the risk of recurrence following local treatment of prostate cancer and is used to make more informed decisions regarding their treatment options.
In the present study, we hypothesized that preoperative PI-RADS v2.1 scoring system would improve the accuracy of predicting postoperative BCR by modifying the existing risk classifications such as the D’Amico classification system. Herein, we explored the clinical utility of preoperative PI-RADS v2.1 score and established a novel risk classification for better prediction of clinical outcomes in patients undergoing RP.
Patients and Methods
Patients. Clinical data from 294 patients who underwent RP at Chiba University Hospital and affiliated institutions between 2006 and 2018 were reviewed. Enrolled patients had undergone prostate needle biopsies and diagnosed with prostate adenocarcinoma with GS classification by the pathologists. All patients underwent RP without neoadjuvant hormone therapy using open, laparoscopic, and robot-assisted procedures. The present study was approved by all institutional review boards and informed consent was obtained from all patients.
Data collection. We investigated the following clinical data for each patient: age at operation, initial PSA (iPSA) level prior to biopsies, prostate volume, biopsy GS sum (bGS), clinical TNM classification, and pathological outcomes of the prostate specimen. The following method was used to measure tumor volumes of prostatectomy specimens (11). All specimens were sectioned transversely at 5-mm intervals and submitted as whole sections. If multiple tumors were present, only the index tumor was measured. All slides containing cancer lesions were imported into ImageJ (National Institutes of Health, Bethesda, MD, USA). Tumor volume was determined by scanning the specimen sections and analyzing the area of the tumor using ImageJ. The following formula was used: total tumor volume (ml)=tumor area × specimen thickness × 1.1 (shrinkage corrected) (11).
MRI protocol and PI-RADS v2.1 scoring system. All patients underwent MRI of the prostate at 3T prior to prostate biopsy. MRI was performed using T1-weighted, T2-weighted, and diffusion-weighted imaging (DWI) sequences to produce an apparent diffusion coefficient map. A high b value (b=2,000) was used for DWI. MRI consisted of T2-weighted images and DWI. Both the bi-parametric MRI (bp-MRI) comprising T2-weighted imaging and DWI, and the apparent diffusion coefficient map were employed by the radiologist to determine the PI-RADS v2.1 score.
PI-RADS v2.1 scores were assessed by the radiologist with non-contrast bp-MRI. The score for each patient was documented using the PI-RADS v2.1 scoring method (5-point scale). The modifications implemented in PI-RADS v2.1 were the scoring of DWI in all zones in categories 2-3 and the revised scoring of the overall rating category in the transition zones (TZs). A DWI score of 4 or 5 elevated the overall PI-RADS rating category from 2 to 3 for lesions with a T2W score of 2 in a TZ (12).
Definition of biochemical recurrence (BCR). The Prostate Cancer Clinical Trial Working Group 2 (PCWG2) definition was employed to determine BCR in the present study (13). BCR was defined as a PSA concentration ≥0.2 ng/ml following RP, measured on two consecutive occasions at least 2 weeks apart. The date of surgery was defined as the date of BCR if PSA level was ≥0.2 ng/ml even postoperatively.
Statistical analysis. Student’s t-test and the χ2 test were used for comparisons between groups. Kaplan–Meier methods (log-rank test) and Cox proportional hazard models were implemented to evaluate the clinical outcomes and predictive factors. Multivariate analysis was performed with clinical parameters showing statistical significance in univariate analyses. JMP Pro 15 (SAS Institute, Tokyo, Japan) was used to for statistical analyses. Statistical significance was set at the level of p<0.05.
Results
Patients background. The study included 294 patients with localized prostate cancer. Median age and PSA level at operation were 67 years old and 7.68 ng/ml, respectively (Table I). Pre-biopsy PI-RADS v2.1 score was 44.9% for score 4 and 24.8% for score 5. Sixty-seven patients (22.8%) were diagnosed with bGS ≥8 prior to surgery. A positive resection margin (RM) was observed in 31.2% of patients. Median tumor volume was 1.88 ml and 17% showed BCR following RP. Median observation period in this study was 63.43 months (Table I).
Patient background.
The prognostic significance of PI-RADS v2.1 scoring system and its relation to patient backgrounds. We divided patients into those with PI-RADS v2.1 score ≤3 and those with ≥4 and compared patient backgrounds between groups (Table II). Patients with a score ≥4 were more likely to be older (p=0.0465), with a higher iPSA level (p=0.0006), more advanced T stage (p=0.0002) and classification as high risk by the D’Amico classification (p=0.0033). A higher incidence of a positive RM was observed in patients with a score ≥ 4 as compared to those with a score of ≤3 (39.5 vs. 12.6%, p<0.0001). In addition, higher PI-RADS v2.1 score was correlated positively with larger tumor volume (p<0.0001, Table II and Figure 1A). Median tumor volumes were 0.57 ml, 0.78 ml, 0.73 ml, 1.87 ml, and 4.28 ml for scores 1-5, respectively (Figure 1A). Kaplan–Meier analysis showed shorter progression-free survival (PFS) in patients with PI-RADS v2.1 score 5 than those with ≤4 (p<0.0001, Figure 1B). Furthermore, to moderate the difference in patients’ backgrounds, the propensity score-matching (PSM) method was employed for evaluating the prognostic value. After 1:1 PSM based on age at operation, initial PSA, and bGS, a total of 112 patients, 56 each, were considered. Patients with preoperative PI-RADSv2.1 score 5 had unfavorable outcomes as compared to those with ≤4 (p=0.0366, Figure 1C).
Characteristics of patients classified according to the prostate imaging-reporting and data system (PI-RADS) v2.1 score.
Prognostic significance of the PI-RADS v2.1 score in patients who underwent radical prostatectomy. (A) Tumor volumes calculated from prostatectomy specimens at each PI-RADS score. (B) Kaplan–Meier analysis classified by the PI-RADS score ≤4 vs. 5 for progression-free survival (PFS). (C) Kaplan–Meier analysis classified by the PI-RADS score ≤4 vs. 5 for PFS after propensity score-matching.
In addition, we investigated a prognostic significance of PI-RADS v2.1 score 5 using Cox proportional hazard models and found that iPSA [hazard ratio (HR)=1.88, p=0.0291], PSA density (HR=1.09, p=0.0008), clinical T stage (≥2c) (HR=4.16, p<0.0001), bGS (≥8) (HR=4.37, p<0.0001), and PI-RADS v2.1 score 5 (HR=4.12, p<0.0001) were associated with PFS in univariate analyses (Table III). After multivariate analysis, clinical T stage [HR=2.32, 95% confidence interval (CI)=1.23-4.38, p=0.0093], bGS (HR=2.81, 95%CI=1.55-5.1, p=0.0007), and PI-RADS v2.1 (HR=2.24, 95%CI=1.19-4.23, p=0.0124) were found to be independent prognostic factors for PFS (Table III).
Uni- and multivariate cox proportional hazard models for progression-free survival.
Prognostic value of the D’Amico classification and PI-RADS v2.1 scoring system. Based on the prognostic importance of PI-RADS v2.1 score, we hypothesized that PI-RADS v2.1 could improve the capability to predict clinical outcomes of existing risk classifications (e.g., D’Amico classification). We confirmed that the D’Amico classification system including clinical T stage, PSA level, and bGS stratified patient prognosis in our cohort (Figure 2A). Patients with the D’Amico high-risk had shorter PFS than those with intermediate risk (p<0.0001, Figure 2A). Of note, PI-RADS v2.1 score 5 significantly differentiated prognosis in both intermediate- and high-risk groups in the D’Amico classification (p=0.0174 and p=0.0013, respectively, Figure 2B). Moreover, patients with intermediate-risk and PI-RADS v2.1 score 5 showed comparable prognosis as compared to those with high-risk and PI-RADS v2.1 score ≤4 with 84.8% and 78.1% in 5-year PFS rate (p=0.8364, Figure 2B). HR (high vs. others) of the risk classification was improved from 4.74 to 7.2 when combined PI-RADS v2.1 score to the D’Amico classification (Table IV). This analysis indicated the importance of PI-RADS v2.1 score in addition to the D’Amico risk classification for better risk classification when considering treatment options.
Validation of the risk classifications for progression-free survival following radical prostatectomy. (A) The D’Amico classification system. (B) Risk classification integrating the D’Amico classification and PI-RADS v2.1 score.
Novel risk classification to predict BCR following RP. We further proposed a novel risk classification comprising clinical T stage (≥2c), bGS (≥8), and PI-RADS v2.1 score from multivariate analysis (Table III). This novel classification was defined as low (met 0 factors), intermediate (met 1 or 2 factors), or high (met 3 factors) and Kaplan–Meier analysis showed significant difference between these three groups (low vs. intermediate: p=0.0008; intermediate vs. high: p<0.0001) (Figure 3A). The HR for the high-risk group compared to others was 9.71 (95%CI=4.38-19.35) (Table IV).
Prognostic significance of novel risk classifications. (A) A risk classification based on cT ≥2c, bGS ≥8, and PI-RADS 5. (B) A risk classification based on PSA ≥7.68, bGS ≥8, and PI-RADS 5.
Comparison of the capability of risk classifications to predict progression-free survival.
In addition, we established another risk classification including PSA (≥7.68 ng/ml), bGS (≥8), and PI-RADS v2.1 score, since PI-RADS score basically reflects cancer size and might have a positive correlation with clinical T stage. Risk classification was similarly defined, and significant differences were observed between low-, intermediate-, and high-risk groups (p<0.0001 and p=0.0069, respectively) (Figure 3B). The HR was 7.03 (95%CI=2.66-15.5) (Table IV). Thus, our risk classification identified a patient population with extremely poor prognosis, achieving a higher HR than the conventional classification.
Discussion
Our study revealed the prognostic importance of the preoperative PI-RADS v2.1 scoring system in patients who had undergone RP. Higher PI-RADS score was associated with increased tumor volumes in the RP specimens. Furthermore, novel risk classifications integrating PI-RADS v2.1 score and the D’Amico classification were developed to improve prognostic capability in comparison with the D’Amico classification alone.
Previous studies have focused on how this scoring system could improve the accuracy of targeted prostate biopsy and avoid unnecessary biopsies (7, 14-16). Positive biopsy rates were significantly improved by fusion prostate biopsy using PI-RADS 4 and 5 lesions as compared to systematic biopsy (64% vs. 42.9%, respectively; p=0.035) although inclusion of PI-RADS 3 did not improve rates, indicating that PI-RADS ≥4 lesions should be recommended for prostate biopsy (16). Wang et al. established a nomogram including bpMRI PI-RADS v2.1 to minimize unnecessary biopsies (7). They found that a predictive model based on bpMRI could improve the ability of clinically significant prostate cancer detection and bpMRI score was correlated strongly with Gleason grade group (7). These findings indicated the clinical utility of PI-RADS scoring system to determine the relevance of performing prostate biopsies.
In addition, several studies have shown that pre-biopsy PI-RADS score could predict clinical outcomes following definitive local treatment (17, 18). A systematic review and meta-analysis revealed that higher PI-RADS v2 classifications were correlated with an increased risk of BCR after local treatment (18). Gandaglia et al. investigated 804 patients who received prostate biopsies and developed a risk classification for predicting PSA failure following RP (19). A predictive model including PI-RADS v2 score achieved the highest accuracy to predict clinically significant prostate cancer for identifying patient populations harboring a higher risk of early recurrence after operation (19). Furthermore, a recent study proposed simplified PI-RADS (S-PI-RADS) that is based on bi-parametric MRI (bpMRI) and is easier to use in clinical practice (20). S-PI-RADS has been found to enhance the detection and diagnosis of PCa as well as local recurrence following radiotherapy and RP (20). These results suggested a prognostic role for PI-RADS v2 classification in addition to diagnosis prior to prostate biopsies (18). However, few reports have examined the prognostic significance of PI-RADS v2.1 since the revision in 2019.
The D’Amico classification system has emerged in 1998 and has become one of the most widely used modalities for risk assessment of localized prostate cancer (8). The system is based on tumor stage by serum PSA level, Gleason grade, and clinical T-score, and divides patients into low-, intermediate-, and high-risk groups to evaluate the probability of recurrence (8). These risk groups have been employed in determining the duration of androgen deprivation therapy (ADT) when radiation therapy is administered (21-23). Furthermore, this risk classification has also been used to determine whether and to what extent lymph node dissection should be performed, and to determine the course of treatment for RP (24, 25). Mandel et al. studied the rationale of lymph node dissection among the D’Amico intermediate-risk patients, and found that the detection rate for lymph node metastasis was low among patients with GS ≤6, cT ≤2b, PSA 10-20 ng/ml, indicating that lymph node dissection may not be necessary in some intermediate-risk cases (25). Thus, this classification is an indicator that serves as a basis for making treatment decisions in localized prostate cancer.
However, this classification has never been modified, and its ambiguity is sometimes noted. In particular, the clinical T stage classification has been obscure in that the diagnostic method is not defined as digital rectal examination or MRI image finding (26). In addition, there is a difference between the National Comprehensive Cancer Network (NCCN) guidelines and the D’Amico classification, e.g., cT2c is classified as an intermediate risk in the NCCN guidelines, however, as high risk in the D’Amico classification system. Our findings showed that PI-RADS scoring was significantly correlated with tumor volume. Based on these results, we hypothesized that the PI-RADS scoring system might offer an alternative implement to the cT stage. Our novel risk classification including PI-RADS score, bGS, and PSA level achieved higher HRs than the D’Amico classification. Furthermore, given that the PI-RADS scoring system is an objective indicator and can be clearly scored, it might be a better clinical biomarker of tumor progression in combination with PSA level and bGS.
In conclusion, we indicated the prognostic significance of the PI-RADS v2.1 scoring system in patients who had undergone RP. The PI-RADS v2.1 scoring system can further stratify patient prognosis in the D’Amico classification and can be incorporated into risk stratification schemes to improve the precision of predicting patient prognosis. Our current exploration might help the decision-making for treatment and post-treatment follow-up in patients with localized prostate cancer who underwent definitive treatment.
Acknowledgements
The present study was supported by the KAKENHI grant 23K19497 and Translational Research Grant of Urological Oncology.
Footnotes
Authors’ Contributions
Conceptualization: Y.F., Y.Y., S.S.; Methodology: Y.F., Y.Y., T.H., X.Z., K.S., S.N., Y.K., H.S., Y.G., T.S., Y.I.; Supervision: M.K., A.F., S.S., T.U., T.I.; Writing – original draft: Y.F.; Writing – review & editing: Y.Y., S.S.
Conflicts of Interest
The Authors declare that there are no conflicts of interest in relation to this study.
- Received October 14, 2023.
- Revision received November 10, 2023.
- Accepted November 13, 2023.
- Copyright © 2023 International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved.
