Percentage of Positive Biopsy Cores Predicts Presence of a Dominant Lesion on MRI in Patients with Intermediate Risk Prostate Cancer

Risk strati ication for prostate cancer utilizes well-established parameters of T stage, Gleason score, and prostate-speci ic antigen (PSA) level [1]. More recently, additional factors such as the percentage of positive biopsy cores (PPCs), percentage of cancer volume (PCV), and maximum involvement of biopsy cores (MIBC) have been shown to have prognostic value, particularly in National Comprehensive Cancer Network (NCCN) intermediate risk patients [2-5]. Magnetic resonance imaging (MRI) of the prostate has been shown to have prognostic signi icance [6-9] and has been shown to correlate with various prognostic factors seen on biopsy [10-13]. These studies have evaluated prognostic factors and MRI indings of extracapsular extension (ECE) and Abstract


Introduction
Risk strati ication for prostate cancer utilizes well-established parameters of T stage, Gleason score, and prostate-speci ic antigen (PSA) level [1]. More recently, additional factors such as the percentage of positive biopsy cores (PPCs), percentage of cancer volume (PCV), and maximum involvement of biopsy cores (MIBC) have been shown to have prognostic value, particularly in National Comprehensive Cancer Network (NCCN) intermediate risk patients [2][3][4][5]. Magnetic resonance imaging (MRI) of the prostate has been shown to have prognostic signi icance [6][7][8][9] and has been shown to correlate with various prognostic factors seen on biopsy [10][11][12][13]. These studies have evaluated prognostic factors and MRI indings of extracapsular extension (ECE) and seminal vesicle invasion (SVI); however, the relationship of these prognostic factors with the inding of a dominant lesion on MRI has not been yet investigated, nor has the comparison of PPC, PCV, and MIBC with MRI indings. Therefore, we assessed the role of PPC, PCV, and MIBC on not only MRI indings of ECE and SVI, but also on the presence and number of dominant lesions. We focused speci ically on intermediate-risk patients.

Methods
Sixty-ive patients with clinically localized intermediate-risk category prostate cancer, who had undergone high-ield-strength pretreatment MRI scans performed during a prede ined time frame (2007-2011), were selected for a retrospective cohort study. Approval for this study was obtained from the Loma Linda University Institutional Review Board. Patients had to have pathology reports with information necessary to calculate PPC, PCV, and MIBC, including the percentage of cancer reported in each core and the total number of biopsy cores taken. These parameters have been previously de ined [2][3][4]14]. Patients starting androgen deprivation therapy prior to undergoing prostate MRI were excluded.

MRI
Images were acquired using either a 1.5 Tesla or 3 Tesla MRI unit. Patients were imaged in the supine position. Patients who underwent imaging with 1.5 T did so with an endorectal coil; those with 3 T imaging did so without an endorectal coil. Cases were reviewed by two body trained radiologists with, respectively, 22 years and ive years of experience. Readers were aware that the patients had prostate cancer as per routine, but were blinded to other clinical information such as PSA and Gleason score. Reader consensus was obtained at the time of readout regarding presence of extracapsular extension, seminal vesicle invasion, and presence of disease in each sextant by T2 and advanced diffusion coef icient (ADC) imaging, including the presence of one or more dominant lesions. The term, "dominant lesion," has been previously de ined with some variability [7,12,[15][16][17], we de ined it as a moderately well-de ined focus of T2 hypointensity within the prostate peripheral zone.

Statistical analysis
Results were analyzed for association between these indings on MRI and the results of clinical and pre-treatment prostate biopsy data, including PSA, Gleason Score, T stage, PCV, PPC, and MIBC. Statistical analysis was performed utilizing independentsamples t-test, Spearman's correlation, logistic regression, and contingency tables. A receiver-operator characteristic (ROC) analysis was done in an effort to de ine clinical cutoffs for PPC, PCV, and MIBC. Analysis was performed using SPSS statistical software.

Results
Sixty-ive patients were identi ied for this study. One was excluded because he was found to be in the NCCN low-risk category and two were excluded because they had started androgen deprivation therapy prior to undergoing prostate MRI, leaving 62 patients available for analysis. Clinical and pathologic characteristics are listed in table 1.

MRI Findings
Complete MRI indings are listed in table 2. Thirty-nine patients (63%) were found to have at least one dominant lesion on MRI. Twenty-four patients (39%) were found to have stage T3 disease on MRI, four patients had extracapsular extension only, 14 patients had seminal vesicle invasion alone, and six patients had both ECE and SVI. Two patients were found to have node-positive disease.

Variables associated with MRI fi ndings
Established prognostic factors of PSA, T Stage, and Gleason score were not found to be statistically signi icant predictors of MRI indings, including the presence of  Next, receiver-operator characteristic (ROC) analysis was done in an effort to de ine clinical cutoffs for PPC, PCV, and MIBC, which could be utilized for predicting a dominant lesion. Again, all three variables were found to predict the presence of a dominant lesion (Area under the curve 0.71, p=0.005; 0.69, p=0.011; and 0.67, p=0.028 respectively). PPC of >=50%, PCV >=15% and MIBC of >=50% were determined to be optimal cut points based on ROC analysis. Contingency tables were constructed (Table  3) and chi-square analysis was done using these cutoffs.
On multivariate analysis, only PPC remained a statistically signi icant predictor of a dominant lesion on MRI, whereas PSA, Clinical T-stage, Gleason score, PCV, and MIBC were not signi icant predictors of such a lesion (Table 4).

Discussion
Intermediate risk prostate cancer patients comprise a heterogeneous population in terms of survival. Others have attempted to additionally stratify intermediate risk patients by using various clinical factors such as a primary Gleason score of 4, PPC >=50%, and the presence of multiple intermediate risk features [18]. Because of the  heterogeneous nature of this patient population, however, there is a need to further assess potential differences in patients within this intermediate-risk group in an effort to ultimately optimize patient care.
In this study of intermediate-risk prostate cancer patients we found that the risk of having a dominant lesion on MRI was signi icantly associated with the percentage of positive biopsy cores. Further, this risk is independent of other prognostic factors such as PSA, T stage, and Gleason score. These indings are particularly relevant to radiation therapy for prostate cancer. Given current ongoing interest in delivering additional radiation dose to MRI dominant lesions in patients with prostate cancer [17,[19][20][21], our study provides additional data on the intermediate-risk patients who are more likely to harbor a dominant lesion and thus be candidates for this type of treatment.
The concept of a dominant MRI lesion has been described in the literature utilizing terms such as "dominant intraprostatic lesion," "dominant tumor," and "index lesion" [7,12,[15][16][17]. Given that the exact de inition of a dominant lesion has had some variability, in our study we de ined it as a moderately well-de ined focus of T2 hypointensity within the prostate peripheral zone. This typically had a gradient of demarcation with adjacent more normal-appearing gland. By contrast, any area of ill-de ined T2 hypointensity within a prostate sextant was designated as having probable tumor in iltration without a nodular morphology and was classi ied as simply T2 hypointense.
To our knowledge, this study is the irst to associate pathologic data from prostate biopsies with the presence of a dominant lesion on MRI. Previous studies have looked at various pre-clinical treatment factors and MRI indings. One study showed that PPC and perineural invasion correlated with T3 disease on MRI; however, the cohort of patients in that study included low-and high-risk groups [10], while our study focused speci ically on intermediate-risk patients. Studies have attempted to integrate MRI into models utilizing biopsy data to predict more extensive disease on pathology after prostatectomy [6,11]. Studies have also looked at MRI indings and clinical outcomes after externalbeam radiation therapy (EBRT). One publication showed that index lesions, among other indings, were associated with worse outcomes after EBRT; however, the authors did not correlate the clinical predictors of PPC or PCV to MRI indings and outcome data [7].
Our study has some discrepancies with previously published data. First, one previous study showed that PPC predicted for the presence of T3 disease on MRI [10], a inding we did not demonstrate. One possible explanation for this is the difference in patient selection, as our study was limited to intermediate-risk patients, while the aforementioned study included a large number of low-risk patients and some high-risk patients. Another study showed that patients with SVI on MRI had higher Gleason scores and pretreatment PSAs [8], which contrasts with our indings. Again, that study included patients in low-and high-risk groups, thus emphasizing the importance of interpreting the present study in the context of intermediate risk patients.
Our study has several indings that generate hypotheses in both the pathology and the clinical management of prostate cancer. Clinical studies have shown PPC as an independent predictor of outcomes after EBRT [3.14], in our study, we have shown that PPC predicts for presence of a dominant nodule on MRI. The presence of a dominant nodule on MRI may be the reason why PPC predicts for outcomes after EBRT. A recent study on MRI-guided biopsies of these dominant lesions showed that Gleason scores of these biopsied lesions are higher than Gleason scores obtained on routine biopsy [22]. These recent indings about the underlying pathology of MRI dominant lesions, combined with the data presented in this study, may lead to further investigation of the biology of these dominant lesions and their role on the natural history of prostate cancer.
Our indings propose many hypotheses of clinical importance as well. Several clinical questions can be asked, such as: Should all patients with PPC >=50% undergo MRI? Should intermediate-risk patients with dominant lesions on MRI undergo additional guided biopsies in attempt to accurately assess their Gleason score? Or should patients with dominant lesions undergo an additional radiation boost? These potential questions could be answered by future prospective studies.
This study has limitations. It is a retrospective investigation from a single institution. Another limitation is the type of imaging used. In our study, most patients underwent scans on 1.5T MRI units obtaining T2 and ADC sequences with use of an endorectal coil; however, many scans done today are multiparametric, performed on 3T MRI units, which includes additional sequences to further characterize lesions. These other parameters were not assessed in this study.
In conclusion, we were able to demonstrate a strong association of PPC and the presence of a dominant lesion on MRI in intermediate-risk patients, independent of PSA, Gleason score, and T stage. These indings suggest that PPC may help predict presence of a dominant lesion on MRI and thus help identify patients who would be candidates for radiation dose escalation treatment to the dominant lesion and/or selective biopsies of the dominant lesion. Given the retrospective nature of this study, these indings are hypothesis-generating and may guide the design of future prospective studies.

Summary
Percentage of positive biopsy cores, percentage of cancer volume, and maximum involvement of biopsy cores each have been shown to have prognostic value and correlate with magnetic resonance imaging indings of extracapsular extension and seminal vesicle invasion. However, the relationship of these prognostic biopsy factors to MRI indings of the presence of a dominant lesion, has not yet been investigated. Accordingly, we included 65 patients with intermediate-risk prostate cancer in a retrospective cohort, seeking to correlate these prognostic biopsy factors to MRI outcomes. Findings of extracapsular extension, seminal vesicle invasion, and presence and number of dominant lesions were noted. Statistical analyses were performed. We determined that patients with one or more dominant lesions on MRI had a signi icantly higher mean percentage of positive biopsy cores, percentage of cancer volume, and maximum involvement of biopsy cores than those without a dominant lesion on MRI. Multivariate analysis revealed that only percentage of positive biopsy cores remained a statistically signi icant predictor for a dominant lesion on MRI; prostate-speci ic antigen, clinical T-stage, Gleason score, percentage of cancer volume, and maximum involvement of biopsy cores were not signi icant predictors of a dominant lesion. We concluded that percentage of positive biopsy cores is a signi icant predictor for the presence of a dominant lesion on MRI. This hypothesis-generating inding should be con irmed in a prospective trial.