Characteristics and Oncological Outcome of Clinical T3a Prostate Cancer Patients Undergoing Radical Prostatectomy in the Multi-Parametric MRI Era

We aimed to reveal the association between the method of diagnosis (multi-parametric magnetic resonance imaging [mpMRI] and digital rectal examination [DRE]) and oncological outcomes of patients with clinical T3a (cT3a) prostate cancer after radical prostatectomy (RP) and stratify them according to oncological risk. We included 132 cT3a prostate cancer patients who underwent RP between 2008 and 2018. The biochemical recurrence (BCR)-free survival rate was evaluated according to the method of diagnosis (mpMRI alone; mpMRI group vs. DRE [with or without mpMRI]; DRE group). Several preoperative factors were evaluated in the multivariate analysis. Patients were divided into risk groups by our prediction model. The mpMRI group had signicantly longer BCR-free survival than the DRE group (p<0.0001). The method of diagnosis (hazard ratio [HR]=2.69; 95% condence interval [CI] 1.45-5.06; p=0.0017) and % positive cores (HR=4.36; 95% CI 1.14-16.5; p=0.031) were independent prognostic factors. Patients were divided into three risk groups based on these factors. There was a signicant difference in BCR-free survival rate among the groups (p=0.0002). The method of diagnosis of cT3a prostate cancer was associated with BCR-free survival, and we categorized patients into risk groups. These assessments were attributable to the appropriate therapeutic strategy for patients with cT3a prostate cancer.


Introduction
Clinical T3a (cT3a) prostate cancer is present in 10-40% of all newly diagnosed prostate cancer, and they are heterogenous in terms of pathological T (pT) stage, biopsy Gleason score (GS), and initial prostate speci c antigen (PSA) level. [1][2][3][4] The oncological outcomes of patients with cT3a prostate cancer are also variable. [1][2][3][4] Available treatment strategies for cT3a prostate cancer are selected based on clinical characteristics as well as personal preference. Several studies have compared oncological outcomes and complications after treatment, but there is no clari ed protocol for a multimodal treatment strategy for cT3a prostate cancer. [5][6][7] As some patients with cT3a prostate cancer are at high risk of symptomatic progression, patients with cT3a prostate cancer should be divided into multiple oncological risk groups and undergo appropriate management. Establishing an oncological risk classi cation system for cT3a prostate cancer is one of the most signi cant challenges in prostate cancer treatment at present.
Although the diagnosis of cT stage in prostate cancer was conventionally based on the results of digital rectal examination (DRE), multiparametric magnetic resonance imaging (mpMRI) is currently being used for more accurate preoperative diagnosis. The diagnostic accuracy of DRE to detect local invasion and predict patients' prognosis is limited because locally invasive tumors are often under-staged. 8 In contrast, the positive predictive value (PPV) of mpMRI for extracapsular extension and the negative predictive value of mpMRI for seminal vesicle invasion are relatively high. 9 Advancements in mpMRI and development of the Prostate Imaging Reporting and Data System have improved the accuracy of prostate cancer detection and cT staging. [10][11][12] However, whether the method of diagnosis can be used to predict oncological outcomes remains unclear.
The aim of this study was to divide patients with cT3a prostate cancer into several risk groups and predict their oncological outcomes accurately in order to identify appropriate therapeutic strategies. We evaluated some preoperative characteristics, such as the method of diagnosis and mpMRI ndings, made a clear distinction between this study and previous studies which focused on pathological ndings.

Patients characteristics
The preoperative characteristics of the patients in this study are shown in Table 1. Out of the 132 eligible patients with cT3a prostate cancer, 38 (28.8%) patients had anterior cancer, 28 (21.2%) had lateral cancer, and 66 (50%) had posterior cancer. There were 89 patients diagnosed using mpMRI alone (mpMRI group) and 43 patients diagnosed using DRE with or without mpMRI (DRE group). There was no signi cant difference in age, GS, initial PSA, or follow-up period between the two groups; however, there were signi cant differences in the site of the tumors and % positive cores. Out of the 66 patients with cT3a posterior cancer, 31 (47%) patients were diagnosed with cT3a on mpMRI in whom DRE did not show cT3a prostate cancer. The remaining 35 (53%) patients were diagnosed with cT3a prostate cancer based on the results of DRE. There was no signi cant difference in age, initial PSA, follow-up period, or % positive cores between the two groups.

Pathological ndings
In the entire cohort, 60.6% of patients in the mpMRI group and 69.8% of patients in the DRE group had pT3 stage cancer. There was no signi cant difference between these two groups (p = 0.21). However, the percentage of pT3 patients was signi cantly higher in patients diagnosed using both mpMRI and DRE (80.7 %; p = 0.03). Out of the 66 patients with cT3a posterior cancer, the percentages of pT3 patients in the mpMRI group and the DRE group were 83.9 % and 68.6 %, respectively. There was no signi cant difference between the two groups. Pathological ndings from radical prostatectomy (RP) specimens are shown in Table 2.

Oncological outcomes
In the entire cohort, the 5-and 10-year BCR-free-survival rates of the mpMRI and DRE groups were 74.7% and 65.6%, respectively, in the mpMRI group, and 63.3% and 27.5%, respectively, in the DRE group. The BCR-free survival of the mpMRI group was signi cantly higher than that of the DRE group (Figure1 (a), p < 0.0001). Among the patients with posterior prostate cancer, the 5-and 10-year BCR-free-survival rates of the mpMRI and DRE groups were 67.4% and 57.8%, respectively, in the mpMRI group, and 36.0% and 25.7%, respectively respectively, in the DRE group. The BCR-free survival of the mpMRI group was signi cantly higher than that of the DRE group (Figure 1(b), p = 0.013).

Prediction models for BCR
In the receiver operating characteristic (ROC) curve analysis, estimation of the area under the curve (AUC) determined the cut-off point for % positive cores as 25%. The sensitivity, speci city, and PPV were 89.1%, 23.3%, and 38.3%, respectively. The patients were divided into three groups based on the method of diagnosis and % positive cores according to the results of the multivariate analysis. The low-risk group included patients diagnosed by mpMRI alone and whose % positive cores were less than 25%. The intermediate-risk group included patients diagnosed by mpMRI alone and whose % positive cores was more than 25% and patients diagnosed by DRE (with/without mpMRI) and whose % positive cores was less than 25%. The high-risk group included patients diagnosed by DRE (with/without mpMRI) and whose % positive cores was more than 25%. The 5-and 10-year BCR-free-survival rates of the low-, intermediate-, and high-risk groups were 86.4% and 86.4%, 68.2% and 57.7%, and 38.1% and 28.6%, respectively. There was a signi cant difference in the BCR-free survival rates among the three groups ( Figure 1(c), p = 0.0002).

Discussion
First, patients with cT3a prostate cancer had heterogenous oncological outcomes after RP and can be divided into several oncological risk groups. Heterogeneity of high-risk prostate cancer has been revealed, and the previous study presents a strati cation of patients with high-risk prostate cancer into oncological risk subgroups based on biopsy GS and initial PSA as well as cT stage. 13 The management of cT3a prostate cancer, including RP, should be determined appropriately depending on their prognostic factors because of their risk of BCR and metastasis. 1 RP with extended pelvic lymph node dissection as part of a multimodal treatment strategy for cT3 prostate cancer is strongly recommended in the EAU prostate cancer guidelines. 14 A previous study reported that RP as a primary treatment for cT3-4 prostate cancer improved cancer-speci c survival and overall survival compared to upfront radiation therapy with androgen deprivation therapy. 15 Furthermore, salvage radiation therapy after RP as a primary treatment strategy decreases the risk of urinary incontinence and erectile dysfunction compared to salvage RP after radiation therapy as a primary treatment. 7 With the recently expanded indication for RP, preoperatively predicting a patient's risk of BCR and metastasis is essential for the treatment strategy. We successfully divided patients into three oncological risk groups, using % positive cores and the method of diagnosis, and observed that the BCR-free survival rates were signi cantly different among these three groups. Thus, the identi ed risk classi cation can be used as a criterion to inform the therapeutic strategy.
Second, the method of diagnosis was the independent prognostic factors of oncological outcomes. We cannot detect anterior cancers by DRE, and patients with posterior cancer accounted for 81.4% of the DRE group in this study. It has been reported that transitional zone/anterior cancers diagnosed by RP specimens are commonly seen in Japanese cohorts; however, anterior cancers are less aggressive. 16 Additionally, the mpMRI has improved the chances of detecting local invasion whose prognosis are better. Previous studies reported that the maximum joint sensitivity and speci city for detecting local invasion using mpMRI reached 71%. 9 In clinical elds, mpMRI is usually used in combination with DRE for higher diagnostic accuracy. In this study, the PPV of both mpMRI and DRE when used to diagnose local invasion was signi cantly higher than using either modality alone. Therefore, the method of diagnosis re ects cancer locations and local invasion which affect oncological outcomes. This is the rst study to reveal that the method of diagnosis is associated with oncological outcomes after RP, and to establish a prediction model composed of preoperative factors only.
Our study has some limitations and sources of bias. First, this was a retrospective, non-randomized study. Therefore, the patient groups have differed in several baseline characteristics that were associated with differences in BCR. Second, multiple urologists performed DRE, which may have led to inconsistencies in examination ndings. Finally, we used BCR as the endpoint because the median follow-up period was 4.9 years, which was too short to estimate prostate cancer metastatic-free survival, cancer speci c survival, or overall survival.
In conclusion, the method of diagnosis of cT3a prostate cancer was a strong prognostic factor for BCRfree survival, and we established a prediction model for BCR-free survival including the methods of diagnosis and %positive cores. Using this model, we successfully divided patients with cT3a prostate cancer into three oncological risk groups. Therefore, the prediction model and the risk classi cation have potential in therapeutic decision-making. Future studies should evaluate the accuracy of this oncological risk classi cation and reveal the optimal therapeutic strategy for each oncological risk group.

Materials And Methods
Patients Cancer Institute Hospital of Japanese Foundation for Cancer Research ethics committee review board approved this retrospective single-institution study (study number: 0994) and waived informed consent requirements and all methods were performed in accordance with the relevant guidelines and regulations.
All patients provided written informed consent. A total of 1771 consecutive Japanese patients with localized prostate cancer underwent RP at the Cancer Institute Hospital, Tokyo, Japan between September 2008 and December 2018. Out of these 1771 patients, we reviewed the medical records of 328 patients diagnosed with cT3aN0M0. Out of these 328 patients, 158 patients who received neoadjuvant hormonal therapy before surgery and 78 patients who did not undergo mpMRI to determine the cT stage were excluded from this study. Finally, 132 patients with cT3a prostate cancer who underwent preoperative mpMRI before RP were included in this study. The cT stage of all 132 patients was estimated by experienced urologists using mpMRI and DRE ndings. After RP, 118 patients (89.4 %) were prospectively observed without any adjuvant treatment until PSA failure was con rmed, and 14 patients An experienced urological radiologist (T.Y.) examined the preoperative mpMRI of eligible patients. The diagnostic criteria of cT3a prostate cancer on mpMRI were asymmetry or invasion of the neurovascular bundles, a bulging prostatic contour, an irregular or spiculated margin, obliteration of the rectoprostatic angle, and a tumor-capsule interface greater than 12mm. 21 The site of the extracapsular lesion on mpMRI was divided into three locations (anterior, lateral, and posterior).

Oncological outcomes
The biochemical recurrence (BCR)-free survival rate was evaluated according to the method of diagnosis (mpMRI alone vs. DRE [with or without mpMRI]). The BCR-free survival rate was also evaluated based on the method of diagnosis for patients with posterior extracapsular lesions on mpMRI. After RP, postoperative PSA levels were measured every 3 months during rst 2 years of follow-up, every 6 months during years 3-5, and annually thereafter. BCR was de ned as a postoperative PSA level ≥ 0.2 ng/mL. 22,23 Statistical analysis The differences in clinicopathological variables based on the method of diagnosis were analyzed by Fisher's exact test, the Mann-Whitney U test, and the Wilcoxon signed-rank test. GS was compared using the Wilcoxon two sample test. The proportion of risk category and pT classi cation were compared using the Mann-Whitney U test. Age and PSA were compared using the Wilcoxon signed-rank test. The BCR-free survival rates, based on the method of diagnosis, were estimated using the Kaplan-Meier method, and differences were assessed with the log-rank test. Multivariate analysis was assessed by a Cox proportion hazard model. In multivariate analysis, the method of diagnosis (mpMRI alone vs. DRE [with or without mpMRI]), % positive cores (continuous), initial PSA (continuous), and GS (<8 vs. ≥8) were evaluated as possible prognostic factors. Estimates from receiver operating characteristics (ROC) curve analysisderived area under the curve (AUC) determined the cut-off points of % positive cores (de ned as the ratio of positive cores in a systematic biopsy specimen). All p-values were two-sided. A p-value of less than 0.05 was considered statistically signi cant. Statistical analyses were performed with JMP version 13. Con icts of interest: The authors have no con icts of interest to declare that are relevant to the content of this article.

Availability of data and material
The data that support the ndings of this study are available on request from the corresponding author, [S.Y.]. The data are not publicly available due to their containing information that could compromise the privacy of patients.