Assessment of the Diagnostic Accuracy of Biparametric Magnetic Resonance Imaging for Prostate Cancer in Biopsy-Naive Men

Key Points Question What are the diagnostic accuracy and negative predictive value of novel biparametric magnetic resonance imaging (MRI) in biopsy-naive men in detecting and ruling out significant prostate cancer? Findings In this cohort study of 1020 men who underwent both biparametric targeted and standard transrectal ultrasound-guided biopsies, low-suspicion biparametric MRI had a high negative predictive value (97%) in ruling out significant prostate cancer on confirmatory biopsies. Meaning The biparametric MRI used as a triage test in this study was associated with improved prostate cancer risk stratification and may be used to exclude aggressive disease and avoid unnecessary biopsies in 30% of men with clinical suspicion of prostate cancer, although further studies are needed to fully explore this new diagnostic approach.

INTERVENTIONS All patients underwent bpMRI (T2-weighted and diffusion-weighted imaging) followed by standard transrectal ultrasound-guided biopsies (all men) and targeted biopsies of men with suspicious bpMRI findings.

MAIN OUTCOMES AND MEASURES Suspicion grades of bpMRI, biopsy results, and NPV of bpMRI
were evaluated for detection of or ruling out significant prostate cancer (Gleason score Ն4 + 3 or maximum cancerous core length >50% for Gleason score 3 + 4). We compared the diagnostic performance of standard biopsies in all men vs standard plus targeted (combined) biopsies restricted to men with suspicious bpMRI findings. The reference standard was combined biopsy results from all men.

RESULTS
A total of 1020 men were enrolled, with a median age of 67 years (interquartile range, 61-71 years) and a median prostate-specific antigen level of 8.0 ng/mL (interquartile range, 5.7-13.0 ng/mL). Combined biopsies detected any and significant prostate cancer in 655 of 1020 men (64%) and 404 of 1020 men (40%), respectively. Restricting combined biopsies to men with suspicious bpMRI findings meant 305 of 1020 men (30%) with low-suspicious bpMRIs could avoid prostate biopsies (biopsy in 715 men with suspicious bpMRIs vs all 1020 men who required standard biopsies [70%]; P < .001). Significant prostate cancer diagnoses were improved by 11% (396 vs 351 men; P < .001), and insignificant prostate cancer diagnoses were reduced by 40% (173 vs 288 men; P < .001) compared with our current diagnostic standard, standard biopsies alone in all men. The NPV of bpMRI findings in ruling out significant prostate cancer was 97% (95% CI, 95%-99%).

Introduction
Standard diagnostic transrectal ultrasonography (TRUS)-guided biopsies are offered to men with clinical suspicion of prostate cancer due to elevated prostate-specific antigen (PSA) levels and/or abnormal digital rectal examination results. However, men without prostate cancer undergo unnecessary biopsies because elevated PSA is not cancer specific. Given the high false-positive rate of PSA, its use for screening purposes is controversial and an area of continuous debate within the medical and urological communities. 1 As standard biopsies are prone to sampling errors because of difficulties in prostate cancer target identification on TRUS, clinically significant prostate cancer may be missed and insignificant prostate cancer detected by the random untargeted sampling, potentially leading to overdetection and overtreatment. 2 In addition, biopsies are invasive and may lead to patient anxiety and morbidity. 3 These limitations have highlighted the need for better diagnostic tools, such as risk calculators, biomarkers, or imaging techniques, 4 to improve selection of men with increased risk of significant prostate cancer who require diagnostic biopsies and subsequent treatment from the proportion of men with either a benign condition or an insignificant prostate cancer that can be managed with expectancy. However, risk calculators are highly influenced by the population studied and newer biomarkers can be costly, may be limited by availability, and have not yet been proven to have the desired level of accuracy in biopsy-naive men with prostate cancer. 5,6 Accurate methods that improve detection of significant prostate cancer while minimizing overdetection and unnecessary biopsies by reducing the number of false-positive results are highly warranted. Growing evidence supports the use of multiparametric magnetic resonance imaging (mpMRI) to solve this problem. 7,8 Magnetic resonance imaging-guided biopsies (targeted biopsies) can be targeted toward the most aggressive part of suspicious lesions detected by mpMRI, improving the detection of significant prostate cancer compared with standard biopsies alone. [9][10][11][12] Conversely, low-suspicion mpMRI may noninvasively exclude the presence of aggressive disease. 13 Accordingly, mpMRI could potentially be used as a triage test to identify biopsy-naive men with clinical suspicion of prostate cancer who might safely avoid unnecessary biopsies.
However, guidelines on prostate MRI 14-16 recommend a full mpMRI prostate examination that includes several anatomical and functional scan sequences as well as intravenous contrast media. This is time-consuming (approximately 40 minutes), expensive, and might be difficult to implement on a large scale. Over time it has become evident that contrast-enhanced imaging and multiple imaging planes often do little to improve the overall clinical picture, especially in the detection and localization of significant prostate cancer. In contrast, a rapid and simple biparametric MRI (bpMRI) method that uses fewer scan sequences and no intravenous contrast media might decrease image acquisition time (approximately 15 minutes) and costs, while retaining sufficient diagnostic accuracy to detect and rule out significant prostate cancer in biopsy-naive men. Such a bpMRI protocol could provide a basis for a prostate MRI triage test prior to biopsy. Consequently, this prospective study assesses the diagnostic accuracy of bpMRI in detecting and ruling out significant prostate cancer in biopsy-naive men with clinical suspicion of prostate cancer. We evaluated the clinical significance of detected cancers and assessed whether bpMRI could be used as a triage test to improve the diagnosis of significant prostate cancer and identify patients who could safely avoid unnecessary biopsies.  17 and adhered to the Standards for Reporting of Diagnostic Accuracy (STARD) reporting guideline criteria. 18 The study inclusion criteria required all men to have clinical suspicion of prostate cancer (PSA Ն4 ng/mL [to convert to micrograms per liter, multiply by 1.0] and/or abnormal digital rectal examination results) that warranted a diagnostic prostate biopsy. The exclusion criteria were prior prostate biopsies, evidence of acute urinary tract infections, acute prostatitis, general contraindications for MRI (eg, claustrophobia, a pacemaker, metal implants), and prior hip replacement surgery or other metallic implants in the pelvic area.

Outcome Measures
The primary end points were the diagnostic accuracy and negative predictive value (NPV) of low-suspicion bpMRI findings in ruling out significant prostate cancer in confirmatory biopsies from biopsy-naive men. Secondary end points included the overall prostate cancer detection rate and detection rates of significant prostate cancer and insignificant prostate cancer stratified by biopsy technique. We also evaluated the clinical value of using bpMRI as a triage test prior to biopsies and estimated the proportion of men who could safely avoid unnecessary biopsies based on low-suspicion bpMRI findings.

bpMRI (Index Test) and Image Analysis
Prior to biopsies, bpMRI was performed using a 3-T MRI magnet (Philips Healthcare) with a pelvic- All bpMRI images were reviewed by the same prostate MRI physician (>5 years of experience) blinded to clinical findings. Suspicious lesions were scored on a 5-point scale according the Prostate Imaging Reporting and Data System version 2 (PI-RADSv2) criteria. 15 However, as the bpMRI protocol does not include dynamic contrast-enhanced imaging, scoring of lesions in the peripheral zone relied solely on diffusion-weighted image findings (dominant sequence), and an equivocal score of 3 was not potentially upgraded to a score of 4 due to lack of positive dynamic contrast-enhanced findings.
All patients were graded overall using this modified PI-RADS score according to their likelihood of having significant prostate cancer (1, highly unlikely; 2, unlikely; 3, equivocal; 4, likely; and 5, highly likely). A modified PI-RADS suspicion score of 2 or lower was perceived as a low-suspicion or negative bpMRI scan result. Patients with no suspicious lesions were assigned an overall modified PI-RADS score of 1.

Standard and Targeted Biopsies
Initially, all patients underwent systematic standard biopsies (10-core extended sextant biopsy scheme) according to guidelines. 19 Any suspicious lesion detected by TRUS was sampled as part of the standard biopsy scheme. Standard biopsies were immediately followed by additional targeted biopsies of any bpMRI suspicious lesions (modified PI-RADS Ն3; 1-2 cores/lesion) using 1 of 2 rigid MRI/TRUS image-fusion systems: HI-RVS system (Hitachi; n = 877) and Uro-Nav system (Invivo;

Histopathological Evaluation and Cancer Significance
All biopsy samples were reviewed by the same genitourinary pathologist (>15 years of experience).
For each prostate cancer-positive biopsy core, the location, Gleason score (GS) based on the Prebiopsy bpMRI suspicion (modified PI-RADS) scores were compared with biopsy results using a χ 2 analysis to determine the association between bpMRI suspicion and positive biopsy findings. We compared the diagnostic performances of the following clinical strategies: (1) standard biopsies in all men, (2) standard plus targeted (combined) biopsies restricted to men with suspicious bpMRIs, and (3) combined biopsies in all men, which served as reference standard. Any patient with significant prostate cancer in either standard or targeted biopsies was classified as having significant prostate cancer on combined biopsies. A McNemar test was used to compare prostate cancer detection rates between biopsy strategies in 2 × 2 contingency tables. The sensitivity and NPV for detecting and ruling out any prostate cancer and significant prostate cancer comparing standard biopsies in all men vs combined biopsies restricted to men with suspicious bpMRIs were calculated to assess our primary outcome measures. Furthermore, the clinical value of the biopsy strategies comparing benefits (significant prostate cancer detection) and harms (unnecessary biopsies) were evaluated using net benefit and decision curve analyses. All anayses were 2-tailed and a P value of less than .05 was considered significant. Statistical analyses were performed using SPSS statistical software version 22.0 (SPSS Inc).

Results
A total of 1063 men were prospectively enrolled and 43 were excluded for various reasons (Figure 1).  Table 2).
The bpMRI modified PI-RADS suspicion scores were associated with the biopsy results (P < .001) (eFigure 1 in the Supplement). The diagnostic yield of significant prostate cancer increased at higher modified PI-RADS scores, and there was a significantly lower significant prostate cancer detection rate in men with low-suspicion bpMRI findings compared with men who had highly     The diagnostic performance consisted of standard biopsies in all men (N = 1020), combined (standard plus targeted) biopsies restricted to men with suspicious biparametric magnetic resonance imaging (bpMRI) findings (n = 715), and combined biopsies in all men (reference standard) (N = 1020). Biopsy results were stratified by cancer significance (primary definition). insPCa indicates insignificant prostate cancer; PCa, prostate cancer; PI mod , modified Prostate Imaging Reporting and Data System score; and sPCa, significant PCa. in men with insignificant prostate cancer when bpMRI was used as a triage test did not change markedly. However, for the tertiary definition of significant prostate cancer (GS Ն3 + 4), the detection rate for the comparison between standard and combined biopsies did not reach the level of statistical significance (McNemar test, P = .11). Sensitivities, NPVs, and net benefit with decision curve analyses are compared in Table 4 and eFigure 2 in the Supplement. Furthermore, restricting combined biopsies to men with suspicious bpMRIs compared with performing combined biopsies in all men reduced overdiagnosis of insignificant prostate cancer by 31% (n = 77; 173 vs 250 men).

Discussion
This study showed that a low-suspicion bpMRI had a high NPV in ruling out significant prostate cancer in confirmatory biopsies. The results suggest that bpMRI may be used as a triage test to exclude the presence of aggressive disease and avoid unnecessary biopsies with its inherent complications (severe infection, rectal bleeding, etc). 3,22 Biparametric MRI suspicion scores were associated with prostate cancer detection rates, and performing combined biopsies (standard and targeted) in all men significantly enhanced the detection of significant prostate cancer compared with standard biopsies alone, which is the recommended diagnostic standard approach in biopsynaive men. If combined biopsies were restricted solely to patients with suspicious bpMRIs, only 8 men with significant prostate cancer would have been missed and significantly fewer men (n = 77) with insignificant prostate cancer would have been diagnosed. Therefore, 305 of 1020 men (30%) could have safely avoided biopsies because most of these men had low-risk disease qualifying for surveillance. Reducing overdiagnoses of insignificant prostate cancer compared with standard biopsies (−40%) or combined biopsies in all men (−31%) might also reduce overtreatment. 2 Our findings are consistent with those of the PROMIS study by Ahmed et al 7 that provided level 1b evidence for the diagnostic accuracy of MRI in detecting prostate cancer. Those findings suggested that if mpMRI were used as a triage test, 1 in 4 men might safely avoid prostate biopsies and the diagnostic ratio of significant prostate cancer vs insignificant prostate cancer would be improved. In  Prior studies evaluated the diagnostic accuracy of bpMRI alone, 24 combined with PSA levels, 25 or compared with mpMRI. 26 In a recent study of 161 biopsy-naive men who underwent bpMRI followed by targeted and standard biopsies, Jambor et al 24   to unnecessary invasive biopsies. The net benefit and decision curve analyses in our study showed that restricting biopsies to men with suspicious (modified PI-RADS 3-5) bpMRI lesions achieved the highest clinical value for all threshold probabilities compared with our current practice-standard biopsies in all men. However, at very low biopsy threshold probabilities, the preferable approach is to perform combined biopsies in all men. Assuming that no urologist would routinely carry out a biopsy in a man with less than a 5% risk of significant prostate cancer (equivalent to performing biopsies in 20 men to find 1 additional significant prostate cancer), using bpMRI to determine whether to perform a biopsy achieved the best clinical outcome balancing benefits and harms.

JAMA Network Open | Urology
In general, we should cautiously consider using bpMRI or mpMRI as a triage test to identify individuals who can avoid prostate biopsies. Numerous factors, including image quality, interpretation, and definition of significant prostate cancer including disease prevalence, can affect the performance of targeted biopsies and the NPV of an MRI. A recent meta-analysis found that the median mpMRI NPVs (suspicion score Ն3) for ruling out any prostate cancer and significant prostate cancer were 82% and 88%, respectively. However, these values were strongly influenced by disease prevalence in the populations studied. 13 We found a lower NPV for any prostate cancer (72%) for a modified PI-RADS score of 3 or higher, but a higher NPV for significant prostate cancer (97%), although the definitions of significant prostate cancer differed. Performing MRI can be expensive and time-consuming, and it would be a major challenge for any health care system to systematically use mpMRIs to diagnose prostate cancer before all biopsies. However, our results confirm that a more rapid and simple bpMRI approach is feasible, is sufficient for MRI/TRUS image fusion, and provides an accurate sector map of the prostate for targeted biopsies. It improves prostate cancer detection and risk stratification in biopsy-naive men and maintains the high diagnostic accuracy of mpMRI. 30,31 Kuhl et al 26 found no significant differences in the diagnostic accuracy of bpMRI and mpMRI in 542 men with elevated PSA who underwent repeated biopsies. However, it is important to note that a low-suspicion bpMRI did not unequivocally rule out any prostate cancer. Nevertheless, the key concern in clinical practice is to detect and rule out significant disease while avoiding unnecessary biopsies.

Limitations
Our study had limitations. It was performed at a single center with 1 dedicated MRI physician reading the bpMRIs and 2 highly experienced TRUS operators performing biopsies. As a result, no interreader variability analyses were done. Less experienced readers and operators might not achieve the same diagnostic yield. Further work would be necessary to evaluate variability between experts and nonexperts. Second, all the patients in our study were from a non-PSA-screened population in whom benign reasons for elevated PSA levels (eg, urinary retention, urinary tract infections) had been ruled out before inclusion. This might explain both the rather high prostate cancer detection rate using standard biopsies and the higher median PSA level (8.0 ng/mL) compared with other studies. 7,11,12,24 The diagnostic accuracy and NPVs of bpMRIs might be different in other patient populations. predictive nomograms and risk calculators are based on standard biopsy results with the inherent limitations of standard biopsy.
Despite these limitations, our data provide evidence for the reliability of using low-suspicion bpMRI findings as a noninvasive diagnostic tool to rule out more aggressive prostate cancer and avoid unnecessary biopsies. Although the use of prebiopsy bpMRI and targeted biopsies significantly improve risk stratification and could benefit clinical practice, the cost-effectiveness and long-term health outcomes using MRI have not been fully explored. Follow-up data and the long-term outcomes of these study patients will be assessed in the future. Furthermore, because bpMRI is a new diagnostic imaging approach, further studies are needed to validate our findings and fully explore the role of bpMRI in prostate cancer management before more widespread implementation into clinical practice.

Conclusions
Low-suspicion bpMRI has a high NPV in ruling out significant disease in biopsy-naive men with clinical suspicion of prostate cancer. Furthermore, bpMRI suspicion scores are strongly associated with prostate cancer detection rates and performing biopsies (standard plus targeted) only in men with suspicious bpMRI findings is the preferred approach for improving the diagnostic ratio of significant prostate cancer to insignificant prostate cancer compared with our current diagnostic standardstandard biopsies in all men. Therefore, bpMRI used as a triage test improves risk stratification and allows for 30% of men with clinical suspicion of prostate cancer to safely avoid unnecessary prostate biopsies with their inherent risks. Further studies are needed to fully explore its future role in clinical prostate cancer management.