The diagnostic accuracy of multiparametric MRI for detection and localization of prostate cancer depends on the affected region

Abstract Objectives To determine the diagnostic accuracy of 3T multiparametric magnetic resonance imaging (mpMRI) for detecting and locating prostate cancer (PCa) on Dickinson's 27‐sector map, using histopathology specimens from radical prostatectomy (RP) as the reference standard. Patients and methods The authors studied a continuous series of 140 patients who underwent RP over three consecutive years. Prior to RP, all patients had mpMRI for detection and localization of PCa and further assessment by biopsy. To minimize the potential of disease progression, 25 patients were excluded because the interval between mpMRI and RP exceeded 6 months, which left 115 patients eligible for analysis. The mpMRI findings were reported using the Prostate Imaging‐Reporting and Data System (PI‐RADS) v2, considering PI‐RADS ≥ 3 to indicate PCa. The histopathology findings from RP specimens were graded using the Gleason scoring system, considering Gleason ≥ 6 to indicate PCa. The location of the tumors was mapped on Dickinson's 27‐sector map for both mpMRI and histopathology and compared by rigid sector‐by‐sector matching. Results The cohort of 115 patients eligible for analysis was aged 66.5 ± 6.0 years at RP. Of the 3105 sectors analyzed, there were 412 true positives (13%), 28 false positives (1%), 68 false negatives (2%), and 2597 true negatives (84%). Across the 27 sectors of the prostate, mpMRI sensitivity ranged from 50% to 100% and specificity from 96% to 100%, while PPV ranged from 50% to 100%, and NPV from 91% to 100%. For the anterior prostate, mpMRI had a sensitivity of 80% (CI, 71%‐86%), specificity of 99% (CI, 99%‐100%), PPV of 91% (CI, 83%‐95%), and NPV of 99% (CI, 98%‐99%). For the posterior prostate, mpMRI had a sensitivity of 88% (CI, 84%‐91%), specificity of 98% (CI, 97%‐99%), PPV of 94% (CI, 92%‐96%), and NPV of 96% (CI, 94%‐97%). Overall, mpMRI had a sensitivity of 86%, specificity of 99%, PPV of 94%, and NPV of 97%. Conclusions The accuracy of mpMRI in detecting and locating prostate tumors depends on the affected region, but its high NPV across all sectors suggests that negative findings may not need corroboration by other techniques.


| INTRODUC TI ON
Prostate cancer (PCa) is the most frequently diagnosed cancer in men, accounting for 20% of cancer diagnoses, and is the second most common cause of cancer-related death in this population. 1 Clinically significant PCa (csPCa) does not have a universally agreed definition, although it is most commonly defined histopathologically using the criteria established by either Wolters et al. 2 or the Prostate Imaging-Reporting and Data System (PI-RADS). 3 A positive finding on digital rectal examination (DRE) and/or a prostate-specific antigen (PSA) result of ≥ 4 ng/mL raises the suspicion of csPCa. 4 Multi-parametric magnetic resonance imaging (mpMRI) is increasingly used for noninvasive detection of PCa, as well as its staging and localization. 5 With improved diagnostic accuracy, mpMRI can guide and enhance biopsy planning, 6 as well as inform appropriate treatment options, such as focal ablative therapies 7 and nerve-sparing surgery. 8 To standardize mpMRI reporting, PI-RADS was proposed in 2012, introducing a scoring system to identify and locate prostate tumors, as well as predict the likelihood of csPCa. The recommended sector map, proposed by Dickinson et al., 9 divided the prostate into 27 sectors to facilitate assessment in predefined regions. PI-RADS v2 was published in 2015, introducing the concept of "dominant sequences" to simplify mpMRI evaluation, and updated the 27-sector map to a 39-sector map, separating the central zone of the prostate. 10 The diagnostic accuracy of mpMRI has been investigated against histopathological specimens obtained by transperineal or transrectal ultrasound-guided (TRUS) biopsy, [11][12][13] template mapping prostate biopsy (TMPB) 14,15 and/or radical prostatectomy (RP). [16][17][18][19][20] While each technique for histopathological sampling has its own limitations, 21 RP is often considered the "gold standard," as it provides a definitive evaluation of the prostate gland. 17 Some studies investigating the accuracy of mpMRI against RP specimens described tumor location using prostate anatomical zones (peripheral or transitional), 19 anatomical "levels" (base, midgland and apex) 15 or PI-RADS sector maps. 22 To the authors' knowledge, however, no contemporary study reported the accuracy of 3T mpMRI with regards to the exact localization of tumors stratified using Dickinson's 27-sector map. This study, therefore, aimed to determine the diagnostic accuracy of 3T mpMRI for detecting and locating PCa (PI-RADS ≥ 3) on Dickinson's 27-sector map, using histopathology specimens from RP as the reference standard.

| Patient selection
The authors retrospectively analyzed the records of 140 consecutive patients who underwent RP under the care of the senior surgeon (CHR) between March 2015 and May 2018. Prior to RP, all patients had an initial suspicion of PCa, indicated by PSA ≥ 4 ng/mL and/or positive DRE, followed by detection and localization using mpMRI and further assessment by TRUS or transperineal biopsy, using both a targeted and randomized sampling approach. All mpMRIs, biopsies and RP procedures were performed at the same institution. None of the patients received any treatment for PCa between mpMRI and RP procedures. Twenty-five patients were excluded from the study because the interval between mpMRI and RP exceeded 6 months, to minimize the potential of significant disease progression following mpMRI, leaving 115 patients eligible for analysis. All patients provided written informed consent for the use of their data and images for research and publication purposes, and the study was approved by the institutional review board.

| MRI technique
All prostate mpMRIs were acquired with the patient in the supine position (feet first), using a 3T unit (Achieva, Philips Healthcare, Eindhoven, NL) with an external pelvic phased-array coil (TorsoXL coil, Philips Healthcare, Eindhoven, NL) but without an endorectal coil. An antispasmodic agent (2 mls of 20 mg/mL hyoscine butylbromide; Buscopan®, Boeringer) was administered intravenously, to minimize peristalsis of the bowel and thereby reduce movement artifact on the image. The imaging protocol used was in accordance with the PI-RADS v2 guidelines, with intravenous contrast injection of 0.1 ml/kg gabobenate dimeglumine (MultiHance®, Bracco Eisai, Tokyo, Japan), administered through a peripheral vein at a rate of 4 mL/s. The sequences acquired before contrast injection included axial, sagittal and coronal T2-weighted fast spin-echo (FSE), an axial T1-weighted FSE, axial diffusion-weighted images (DWI) using b0, b100 and b1500 to generate the apparent diffusion coefficient (ADC) map, and a separate high b value DWI (b2000s/mm 2 ). During contrast injection, an axial three-dimensional (3D) FSE dynamic contrast enhanced (DCE) sequence was acquired. After contrast injection, an axial T1-weighted FSE sequence was acquired.

Conclusions:
The accuracy of mpMRI in detecting and locating prostate tumors depends on the affected region, but its high NPV across all sectors suggests that negative findings may not need corroboration by other techniques.

K E Y W O R D S
diagnostic accuracy, localization, magnetic resonance imaging, mpMRI, prostate cancer, PI-RADS, radical prostatectomy

| Imaging analysis
Each tumor within the prostate gland was identified and evaluated by the same radiologist with 15 years' experience in prostate MRI and graded as per PI-RADS v2 to report likelihood of csPCa (1: highly unlikely, 2: unlikely, 3: equivocal, 4: likely, and 5: highly likely).
Therefore, the present study considered PI-RADS ≥ 3 in a given sector to indicate PCa in that sector, and PI-RADS ≥ 4 in a given sector to indicate csPCa in that sector. Each tumor was then mapped onto Dickinson's 27-sector map, thereby assigning a PI-RADS grade to each sector. Tumor-nodes-metastasis (TNM) staging criteria were used to report presence of extra prostatic extension (EPE), according to the American Joint Committee on Cancer (AJCC), 23 with pT3 staging considered positive for EPE.

| Radical prostatectomy histopathology analysis
All patients underwent robotic-assisted laparoscopic RP by two urological surgeons. Histopathologic whole-mount specimens were and ≥ 1.3 cc; (iii) pT stage 3a or greater (EPE); and (iv) nodal metastasis. The present study considered Gleason score ≥ 6 in a given sector to indicate PCa in that sector, and Gleason score ≥ 7 in a given sector to indicate csPCa in that sector.

| Correlation between mpMRI and histopathology
Histopathology findings were used as the reference standard for tumor detection and the findings on mpMRI and histopathology were compared by rigid sector-by-sector matching 18

| Statistical analysis
The accuracy of mpMRI at detecting PCa (PI-RADS ≥ 3 and Gleason ≥ 6) and csPCa (PI-RADS ≥ 4 and Gleason ≥ 7) in each sector was expressed in terms of sensitivity/specificity and positive predictive value (PPV)/negative predictive value (NPV), with 95% confidence intervals (CI). All values were calculated for the 27 sectors, as well as for the anterior prostate, the posterior prostate, and the prostate overall, by summing the numbers of respective TPs, FPs, FNs, and TNs. The accuracy of mpMRI at detecting EPE was also expressed in terms of sensitivity/specificity and PPV/NPV for the entire prostate. Statistical analyses were performed using R version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria).

| RE SULTS
The cohort of 115 patients eligible for analysis had a mean (± standard deviation) age of 66.5 ± 6.0 years (range, 50.9-77.8) Across the 27 sectors of the prostate, sensitivity ranged from 50% to 100% and specificity from 96% to 100%, while PPV ranged from 50% to 100%, and NPV from 91% to 100% (Figures 2 and 3).

| D ISCUSS I ON
Using histopathology from RP specimens as the reference standard to diagnose PCa, mpMRI had variable sensitivity (50%-100%) and PPV (50%-100%) across the 27 sectors of the prostate, while exhibiting excellent specificity (96%-100%) and NPV (91%-100%) in all sectors. The clinical relevance of these findings is that, while mpMRI is not uniformly reliable at ruling out PCa across some sectors of the prostate (with variable probabilities of missing a tumor), mpMRI is F I G U R E 3 Positive predictive value (PPV) and negative predictive value (NPV) of mpMRI for detection of PCa (PI-RADS ≥ 3 and Gleason ≥ 6), for each sector, the anterior prostate, the posterior prostate, and the overall prostate uniformly reliable at ruling in PCa in all sectors (with very low probability of indicating a tumor that is not present). The accuracy of mpMRI in detecting and locating prostate tumors depends on the affected region, with particular variability across the anterior sectors, but its high NPV across all sectors suggests that negative findings may not need to be corroborated by randomized biopsy, and that focal therapies can be considered for some cases as a less invasive alternative to RP. The accuracy of mpMRI for detection of csPCa compared to PCa, remains high for specificity and NPV, while it is reduced for sensitivity and PPV.
The results of the present study showed that mpMRI had the highest sensitivity (100%) in sectors 2a, 15as, 7a, and 5a, followed by 10p (96%) and 2p (95%), while it had the lowest sensitivity (50%) in sectors 1a, 11a, and 3as, followed by 8a (60%) and 14as (62%). It is worth noting that the sectors where mpMRI had the highest and lowest sensitivities were also those with the fewest observations of PCa, and hence, the largest confidence intervals. The sensitivity of mpMRI was lower for the anterior prostate (80%; CI, 71%-86%) than for the posterior prostate (88%; CI, 84%-91%). Other recent studies reported lower sensitivity for the anterior prostate (62.4% and 78.1%) 26,27 , which remains the more challenging region to diagnose not only on mpMRI, but also by DRE and/or TRUS. Lawrentschuk et al 28   This study has a number of limitations, including its retrospective design, as well as the risk of selection bias that is inherent in studying a population of patients who underwent RP, and as such, the sample population does not represent patients that had mpMRI but did not subsequently undergo RP. This may lead to the overestimation of diagnostic accuracy, as the agreement between mpMRI and histopathology results may be "artificially high". 39 Furthermore, inter-rater agreement was not assessed, as mpMRI images were only evaluated by one radiologist, which could result in a lack of quality control, although inter-rater agreement for PI-RADS v2 has already been shown to be moderate to substantial. 40 Finally, there is the potential for mismatch between mpMRI and RP sectors, due to deformation or shrinkage of RP specimens, and the possible misalignment between the axial plane on mpMRI and the sectioning angle used for histopathology F I G U R E 5 Positive predictive value (PPV) and negative predictive value (NPV) of mpMRI for detection of csPCa (PI-RADS ≥ 4 and Gleason ≥ 7), for each sector, the anterior prostate, the posterior prostate, and the overall prostate specimens. While the limitations of this matching method should be acknowledged, the use of RP as the "gold standard" can be considered a strength of this study, as it provides the highest degree of validation.
A new possibility of imaging approach before biopsy, used complimentary to MRI, could be prostate-specific membrane antigen using positron emission tomography (PSMA PET), which could improve diagnostic sensitivity.
The accuracy of mpMRI in detecting and locating prostate tumors depends on the affected region, but its high NPV across all sectors suggests that negative findings may not need corroboration by other techniques.

ACK N OWLED G M ENTS
The authors are most grateful to Mo Saffarini for his assistance with manuscript preparation.

CO N FLI C T O F I NTE R E S T
MM, SR, SR, ISQ, GADB, and CHR have no conflicts of interest. LS, GH, and SRP report personal fees from ReSurg SA, during the conduct of the study.