The Role of Multiparametric MRI and MRI-targeted Biopsy in Detecting Clinically Significant Prostate Cancer in the Community Setting: A Retrospective Study
Kenneth E. Schmanke, MD,1 Jordan J. Stiverson, MD,1 Rosalee E. Zackula, MA,1 Serge G. Srour, DO,2 Matthew P. Lierz, MS,1 Fadi N. Joudi, MD3
1University of Kansas School of Medicine–Wichita, Wichita, KS; 2Department of Diagnostic Radiology, University of Kansas School of Medicine–Wichita, Wichita, KS; 3Wichita Urology Group, Wichita, KS
Multiparametric MRI and the Prostate Imaging-Reporting and Data System (PI-RADS) have emerged as tools to reveal suspicious prostate lesions and MRI-targeted biopsy has shown potential to avoid repeat prostate biopsies and miss fewer significant cancers. This retrospective study sought to assess the differences in diagnostic yield and sampling efficiency between MRI-targeted and standard biopsies in a community urology practice. We concluded that MRI-targeted biopsy was more efficient than a standard biopsy, although neither technique achieved a superior diagnostic yield of clinically significant cancer in our community setting. We recommend that a standard biopsy be performed alongside targeted biopsy.
[Rev Urol. 2020;22(2):57–66]
© 2020 MedReviews®, LLC
From the time the Surveillance, Epidemiology, and End Results (SEER) database began collecting data in 1975, prostate cancer has remained the most common cancer among men in the United States.1 The traditional pathway to diagnosis has been the discovery of an elevated screening prostate-specific antigen (PSA) level or a firm, nodular prostate on digital rectal exam (DRE), followed by a sampling of the prostate tissue for confirmation. Since the arrival of ultrasound into clinical mainstream in the 1980s, the transrectal ultrasound (TRUS) biopsy has remained the standard of care for this tissue sampling. This biopsy technique typically consists of obtaining 10 to 12 tissue cores in a systematic fashion under ultrasound guidance in order to sample the entire prostate.2 However, even with ultrasound guidance there is variability, incomplete sampling, and an element of operator discrepancy to the procedure, resulting in false-negative biopsies in 20% to 24% of cases and at other times overdetection of low-grade cancer that may never manifest clinically.3,4 More extensive techniques, such as the saturation biopsy, have been attempted with the goal of achieving more complete sampling of the prostate but resulted in neither increased cancer detection nor increased grading accuracy of detected cancer.5,6
With underdiagnosis of aggressive disease and overtreatment of indolent disease remaining major challenges of TRUS biopsy, the past decade saw advances in MRI technology as a potential solution, shifting its role from simply staging prostate cancer to an instrument that can be used to rule out clinically significant disease.7 One tool that assisted in expanding the role of multiparametric MRI (mpMRI) from staging to diagnosis was the implementation of the Prostate Imaging Reporting and Data System (PI-RADS), an algorithmic system for the standardization of performance, interpretation, and reporting of mpMRI.8 The emergence of mpMRI coupled with PI-RADS and targeted biopsy techniques has allowed urologists to miss fewer high-risk cancers and better elucidate which patients can safely avoid biopsy, saving them from the cost and morbidity associated with the procedure.9-14
The purpose of this project was to retrospectively compare the diagnostic yield and sampling efficacy of mpMRI-targeted biopsy to the standard TRUS biopsy in in a non-academic setting. We also aimed to assess the positive predictive value (PPV) of the most recent expert consensus update of the PI-RADS assessment tool in detecting clinically significant prostate cancer in this setting. The implications of this study are that MRI-targeted biopsy could replace the standard TRUS biopsy or that mpMRI could prevent unnecessary biopsy of the prostate.
The study was designed utilizing the Standards of Reporting for MRI-Targeted Biopsy Studies (START) guidelines.15
A 17-month retrospective cohort study was conducted using chart review from a single specialty urology practice between April 2017 and Sept 2018. Inclusion criteria were men ≥18 years of age, who underwent mpMRI for clinically suspected prostate cancer (rising or elevated PSA, abnormal DRE, or atypical small acinar proliferation on previous biopsy) or with previously diagnosed prostate cancer. Exclusion criteria were patients who received previous radiation therapy for prostate cancer, who have missing radiology or pathology reports in their chart, and prostates categorized as PI-RADS <3. In patients with multiple lesions, only the highest category PI-RADS lesion was analyzed. None of the patients in this study have been included in any previously published cohorts.
The primary outcomes were to compare the clinically significant cancer yield and sampling efficacy of standard and MRI-targeted prostate biopsies in a community setting. Sampling efficacy of each biopsy method was compared by assessing the proportion of biopsy cores positive for clinically significant cancer and the mean number of cores taken per diagnosis of clinically significant cancer. The secondary outcome was to determine the PPV of PI-RADS categories 3, 4, and 5 for clinically significant prostate cancer in a community setting. Clinically significant prostate cancer was defined as a Gleason score ≥3+4. Insignificant cancer was defined as a Gleason score of 3+3.
Other patient variables of interest included family history of prostate cancer, DRE findings, prostate volume, prebiopsy PSA, PSA density, and lesion zone and location.
Study data were collected and managed using Research Electronic Data Capture (REDCap) electronic data capture tools hosted at the University of Kansas Medical Center. REDCap is a secure, web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for importing data from external sources.16 Because the data in REDCap contained protected health information, institutional review board approval was obtained.
mpMRI was performed using a 1.5 T magnet and a phased-array body coil only. Small field-of-view (20 cm × 27 cm) axial T2 sequences were obtained in anterior-posterior (AP) and left-right (LR) phase-encoding directions with a reconstruction matrix of 384 × 384. Standard large field-of-view coronal and sagittal T2-weighted sequences were also obtained. Diffusion-weighted images were acquired in AP and LR phase-encoding planes and qualitatively analyzed using high b value (b-1400) diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) sequences. Dynamic contrast enhancement was performed using a dynamic T1 3D mpMRI with a temporal resolution of 9s. The data were analyzed qualitatively and quantitatively using a pharmacokinetic map and time-signal intensity plot processed with DynaCAD (Phillips Healthcare Solutions, Franklin, TN). Additionally, a pre–liver acquisition with volume acquisition (LAVA) and 3-min post-LAVA sequence was obtained.
Prostate lesions were categorized by radiologists with approximately 3 to 4 years of experience using PI-RADS version 2 (V2). The radiologists were not blinded to clinical information. PI-RADS category 3 to 5 lesions were marked using DynaCAD 2D freehand application.
Prior to prostate biopsy, an 8 Hz ultrasound probe was inserted in the rectum and the prostate was measured. Utilizing UroNav software image registration (Phillips Healthcare), the operator first obtained multiple, targeted TRUS biopsies. Then, standard, TRUS biopsies were obtained. MRI-targeted cores were potted separately from the standard cores. The operator was not blinded to the location of the lesion on MRI.
Because there is not a practical gold-standard test for prostate biopsy, we utilized the imperfect standard, TRUS biopsy, as our reference test. The sensitivity of MRI-targeted biopsy (standard biopsy) was calculated as the number of positive MRI-targeted biopsies (standard biopsies) results divided by the total number of cancers detected by either test.
The two biopsy methods were compared for equivalence using a two one-sided test (TOST).
The predictive value of PI-RADS V2 for detecting clinically significant disease was assessed by calculating the PPV of each PI-RADS category.
Data were collected from 532 consecutive charts in which patients received an mpMRI of the prostate (Figure 1). After exclusions, the analysis included 230 patients. Of these patients with a suspicious lesion (PI-RADS category ≥3), 168 (73%) received standard and MRI-targeted biopsies (Table 1).
Of the 168 patients with both standard and MRI-targeted biopsies analyzed, the PI-RADS category distribution was 92, 64, and 12 for PI-RADS 3-5, respectively. One hundred and ten (65%) had cancer on their pathology reports, of which 67% was clinically significant and 33% was clinically insignificant.
A mean of 19 and 6 cores were taken per lesion by standard and MRI-targeted biopsy, respectively. Clinically significant cancer was detected in 57 standard and 51 MRI-targeted biopsies with sensitivities of 0.77 and 0.69, respectively (Table 2). Clinically insignificant cancer was detected in 41 standard and 28 MRI-targeted biopsies. Seventeen patients were upgraded from clinically insignificant disease/no cancer on standard biopsy to clinically significant disease by MRI-targeted biopsy (Table 3). Twenty-three targeted biopsies originally classified as clinically insignificant disease/no cancer were upgraded to clinically significant disease by standard biopsy. Ninety-one (54%) of standard biopsies had concordant Gleason grades with MRI-targeted biopsies, whereas 27 (16%) were downgraded by targeted biopsy and 50 (30%) were upgraded. Fifty-eight patients were found to have no cancer on either sampling method.
When compared for equivalence using TOSTs, the MRI-targeted biopsy technique was found to have superior sampling efficiency than the standard technique (Table 4). However, neither technique was found to be superior in detecting clinically significant disease or missing clinically insignificant disease.
The biopsy results according to PI-RADS category are displayed in Table 5. The PPV of detecting clinically significant cancer correlated positively with each PI-RADS category (Table 6, Figure 2).
An optimal biopsy method would be one that maximizes detection of disease whose course could be altered while minimizing detection of disease that will never manifest with clinical symptoms. Furthermore, this method would have high concordance with whole-gland surgical pathology, result in no morbidity to the patient, and be cost effective. The TRUS-guided diagnostic pathway to prostate cancer has been the standard that has brought urologists closest to this goal since the 1980s, although it is currently being challenged by a pathway that is based in MRI technology. Recently, the PRECISION trial showed that an MRI-based diagnostic pathway resulted in fewer men undergoing biopsy, more clinically significant cancers being identified, less overdetection of clinically insignificant cancer, and fewer biopsy cores being obtained when compared with TRUS-guided biopsy.17 This was the first international, multicenter trial to compare these two diagnostic pathways in men who were biopsy naive.
Other results in literature are more mixed in the comparisons of these two diagnostic pathways. In a metanalysis of 1926 men, overall prostate cancer detection did not differ significantly between targeted and standard prostate biopsies, although the targeted biopsies had a higher rate of detection of significant cancer and a lower rate of detection of nonsignificant cancer.12
In our community-based study, the standard biopsy technique had a higher sensitivity for detecting clinically significant cancer. However, this did not reach statistical significance and, therefore, neither biopsy technique differed meaningfully in their ability to detect clinically significant disease or miss clinically insignificant disease. It is worth noting, however, that when used alone, both techniques missed clinically significant disease at similar rates, with 14% of targeted biopsies and 10% of standard biopsies failing to detect clinically significant disease that was found with the other technique. This does not consider likely false-negative biopsies from both methods, as we know from previously published data that standard biopsies have a false-negative rate of 20% to 24%.3 For this reason, we cannot suggest that targeted biopsies be pursued without concomitant standard biopsies.
A question of active research remains: can mpMRI be utilized to avoid biopsy altogether? The PROMIS study tested the diagnostic accuracy of mpMRI and TRUS against a template biopsy and found that 27% of men might have avoided a primary biopsy.9 Additionally, the ASIST and PICTURE studies found negative predictive values of 85% and 91%, respectively, suggesting that in patients with favorable parameters, a negative MRI might be trusted as truly negative.18 Although it was not the focus of the study, the FUTURE trial showed fairly poor predictability by mpMRI.19 Our study excluded 295 individuals who may have otherwise been biopsied based on their elevated PSA or known diagnosis of cancer but were not sampled because their PI-RADS category was <3. However, we cannot be sure that these were false-negative results or that these patients safely avoided a biopsy.
MRI-targeted biopsy can take several forms including MRI-TRUS fusion–targeted biopsy, cognitive registration TRUS-targeted biopsy, or in-bore MRI-targeted biopsy. Recently, the FUTURE trial became the first multicenter, randomized, controlled trial to publish findings comparing these biopsy methods. They found that in men with previously negative biopsies, there were no meaningful differences in the detection rates of clinically significant disease among these different biopsy techniques.19
The previously mentioned PRECISION trial has been praised for its pragmatism, which included allowing non-academic centers to participate in the study. However, we believe our study to be a uniquely retrospective focus comparing MRI-targeted and standard biopsy techniques in a strictly community setting. We see an added importance of publishing results from the community as only 25% of US urologists practice in academic medical centers, yet the vast majority of published data comes from this relatively small subset of practice types.20 There are many differences between these settings, including availability of resources, differing levels of expertise, and budget allowances that may limit generalizability. Novel for a community urology group, the operators in our study met regularly with the reading radiologists to retrospectively discuss cases and improve communication between the two specialties. Our study was further made unique and generalizable to community urologists by total avoidance of an uncomfortable and expensive endorectal coil, and the use of a 1.5T magnet.
An expert consensus update of PI-RADS was published in 2015, PI-RADS V2.21 This update aimed to overcome limitations of the previous guidelines and simplify the algorithm with the goal of a more global adoption of the scoring system. This included taking into consideration the location and size of a lesion and summarizing the findings in a final 5-point score. Since this update, PI-RADS V2 has been studied systemically and has been found to have a good sensitivity and moderate specificity; however, these studies were plagued by heterogeneity.22-24 Our PPVs of 0.32, 0.53, and 0.92 for PI-RADS 3-5, respectively, is consistent with these systemic analyses, suggesting that the PI-RADS scoring system has application even in a community-based setting. In 2019, an even more contemporary update, PI-RADS V2.1 was published with the intention to further refine limitations and inconsistencies of the PI-RADS utilized in this study.25
Our study has limitations. It is small, non-randomized, and retrospective in nature and did not assess any benefits toward clinical endpoints including cancer-specific mortality. Our study also used data from one, single-specialty urology clinic, in which men with prostate cancer were mostly diagnosed and treated by eight urologists. Therefore, our study results may not be representative of other urological practices, particularly large, tertiary-care academic hospitals. We utilized Gleason scoring to define clinically significant versus insignificant disease. The most appropriate definition of clinically significant cancer has yet to be defined and, therefore, is subject to debate and may limit comparison with similar studies. A significant portion (27%) of suspicious lesions (PI-RADS category ≥3) were not biopsied, potentially lending to a selection bias of our study. Lastly, our study population had a particularly small sample size of very high risk (PI-RADS category 5) lesions.
This study showed that although MRI-targeted biopsy of the prostate is more efficient than a standard biopsy, neither technique achieves a superior diagnostic yield of clinically significant cancer in a non-academic setting. Both sampling methods missed a similar rate of clinically significant disease when compared with each other and, therefore, we cannot suggest that targeted biopsy be used without a concomitant standard biopsy in this setting. Additionally, the PI-RADS scoring system had similar PPVs to those published in the literature, suggesting that expert consensus recommendations of PI-RADS have generalizability in the community setting.