ࡱ > !
| bjbj 6f
@ @ @ @ @ T T T 8 D T > ~ B B B B = = = = = = = $ ? B t 8= @ q @ 8= @ @ B B / = " " " @ B @ B .4 " = " " r + T , B [j , 4 = 0 > , C o! T C , , C @ v- " 8= 8= " > C
: Title: Validation of the Dominant Sequence Paradigm and Role of DCE in PIRADSv2
Authors: Matthew D. Greer, BS1,2; Joanna H. Shih, PhD3; Nathan Lay, PhD4; Tristan Barrett, MD5; Leonardo Bittencourt, MD6; Samuel Borofsky, MD7; Ismail Kabakus, MD8; Yan Mee Law, MD9; Jamie Marko, MD10; Haytham Shebel, MD11; Francesca V. Mertan, BS1; Maria J. Merino, MD12; Bradford J. Wood, MD13; Peter A. Pinto, MD14; Ronald M. Summers, MD/PhD4; Peter L. Choyke, MD1; Baris Turkbey, MD1
1Molecular Imaging Program, NCI, NIH, Bethesda, MD
2Cleveland Clinic Lerner College of Medicine, Cleveland, OH
3Biometric Research Program, NCI, NIH, Bethesda, MD
4National Institutes of Health Clinical Center, Imaging Biomarkers and Computer-Aided Diagnosis Laboratory, Radiology and Imaging Sciences, Bethesda, MD
5University of Cambridge School of Medicine, Department of Radiology, Cambridge, UK
6 Universidade Federal Fluminense, Rio de Janeiro, Brazil
7George Washington University Hospital, Washington DC
8Hacettepe University, Ankara, Turkey
9Singapore General Hospital, Singapore
10Walter Reed National Military Medical Center, Bethesda, MD
11Department of Radiology, Nephrology Center, Mansoura University, Mansoura 35516, Egypt
12Laboratory of Pathology, NCI, NIH, Bethesda, MD
13Center for Interventional Oncology, NCI and Radiology imaging Sciences, Clinical Center, NIH, Bethesda, MD
14Urologic Oncology Branch, NCI, NIH, Bethesda, MD
Keywords: PI-RADSv2, mpMRI, Prostate Cancer, DCE
Advances in Knowledge:
PIRADSv2 scores predict cancer detection rates (15.7%, 33.1%, 70.5%, 90.7% for PI-RADS 2, 3, 4, 5, p<0.05 between each).
PIRADSv2 scores demonstrate equivalent cancer detection rates for peripheral and transition zone tumors for clinically significant disease (p>0.05 for PIRADS=3, 4, 5), except PIRADS=2 (23.8% PZ, 3.6% TZ; p=0.021)
DWI outperforms T2W in the peripheral zone, per PIRADSv2 scoring criteria, validating DWI as the dominant sequence in the PZ (OR DWI 3.49 vs T2W 2.45; p=0.008).
Dynamic contrast enhanced imaging per PIRADSv2 scoring criteria provides meaningful assistance to cancer detection for lesions scored PIRADS = 3 and 4 (OR 2.0; p=0.027).
Implications for Patient Care
Prostate cancer diagnosis is limited by over-detection of clinically insignificant cancers and under-detection of clinically significant cancers by ultrasound guided biopsies. mpMRI assists in detecting clinically significant cancers, but is limited by unclear interpretation criteria. We demonstrate that PI-RADSv2 scoring provides efficient scoring criteria to optimize cancer detection rates by radiologists, although dynamic contrast enhanced imaging may be of more assistance than PI-RADSv2 proposes.
Summary Statement
This study affirms the value of the dominant sequence paradigm put forth in PIRADSv2 especially in the PZ and clarifies that DCE MRI, whose role in PIRADSv2 is controversial, adds significant benefit to PIRADS 3 and 4 lesions in the PZ.
Corresponding Author
Baris Turkbey, MD
10 Center Drive, Room B3B85
Bethesda MD 20892 USA
turkbeyi@mail.nih.gov
Phone: 301-443-2315
Fax: 301-402-3191
Abstract
Purpose: PIRADSv2 was recently proposed to standardize mpMRI for prostate cancer. It includes a dominant sequence paradigm and limits the role of DCE. We sought to validate these assumptions using data from a multi-reader study.
Methods: This HIPAA-compliant, retrospective interpretation of prospectively acquired data was approved by the local ethics committee. Patients were treatment-nave who had ERC 3T MRI (T2W, ADC, b2000, DCE) between 4/2012-6/2015. 163 patients were evaluated; 110 with prostatectomy after mpMRI and 53 with negative mpMRI and biopsies. 9 radiologists participated. Lesions were scored with PIRADSv2 and correlated to whole-mount prostatectomy. Cancer detection rates (CDR) for overall, T2W, DWI PIRADS scores were calculated in PZ, TZ. To determine benefit of DCE, logistic regression models were used with PIRADS and DCE scores as independent variables.
Results: 654 lesions (420 PZ) were detected. The CDR for PIRADS=2, 3, 4, 5 was 15.7%, 33.1%, 70.5%, 90.7%, respectively. There was no difference in CDR for PZ and TZ lesions at each PIRADS score except PIRADS=2 (23.8% PZ, 3.6% TZ; p=0.021). DWI outperformed T2W in PZ (OR 3.49 vs 2.45; p=0.008). Lesions with PIRADS DWI=3 and DCE=positive in PZ demonstrated higher CDRs than PIRADS=3 lesions (67.8% vs 40.0%, p=0.02). Addition of DCE to DWI in PZ demonstrated a benefit (OR 2.0; p=0.027) with an increase in CDR of 15.7%, 16.0%, and 9.2% for PIRADS=2, 3, 4, respectively.
Conclusion: PIRADSv2 scoring predicts cancer detection rates with equivalent performance in PZ and TZ. The addition of DCE to DWI scores in the PZ provides meaningful improvements in CDR. Introduction
The Prostate Imaging Reporting and Data System version 2 (PIRADSv2) was introduced recently to promote global standardization and diminish variation in the acquisition, interpretation, and reporting of prostate multiparametric (mpMRI) examinations. It was intended to be a living document based on experience and accrual of scientific data with a strong likelihood for other versions over time ADDIN EN.CITE Radiology2015380(1)380380017American College of RadiologyMR Prostate Imaging Reporting and Data System version 2.02015http://www.acr.org/Quality-Safety/Resources/PIRADS/12/2015(1). Studies to test the efficacy of PIRADSv2 were encouraged to validate its utility.
One major change in PIRADSv2 was the introduction of the concept of the dominant sequence which varies according to location of the lesion. For transition zone (TZ) lesions T2-weighted (T2W) imaging was proposed to be the dominant sequence and for peripheral zone (PZ) lesions, diffusion weighted imaging (DWI) was proposed as the dominant sequence. Dynamic contrast enhanced imaging (DCE) was assigned a minor role, mainly to upgrade PIRADS=3 based on DWI score in the PZ to PIRADS=4 if the DCE was considered positive (i.e. 3+1 lesion).
The dominant sequence paradigm was first proposed by Vache et al ADDIN EN.CITE ADDIN EN.CITE.DATA (2) and was based on studies indicating that DWI and DCE were of limited value in the TZ as BPH nodules can imitate the vascularity and diffusion parameters of cancers ADDIN EN.CITE ADDIN EN.CITE.DATA (3-5). For the PZ, DWI emerged as the dominant sequence with higher sensitivity for cancer. Moreover, lower apparent diffusion coefficient images (ADC) were associated with higher Gleason grades ADDIN EN.CITE ADDIN EN.CITE.DATA (5-7). The limited role of DCE in PIRADS scores reflected current opinion that DCE contributes little to the diagnosis of prostate cancer as it is primarily useful when DWI is not definitive ADDIN EN.CITE ADDIN EN.CITE.DATA (8, 9).
We sought to validate the dominant sequence paradigm and limited role of DCE of the proposed PIRADSv2 scoring system. We utilized data from a multi-reader study consisting of nine readers from eight different institutions and used prostatectomy specimens and negative mpMRI/biopsy controls as the gold standard for pathology.
Materials and Methods
Study Population
This HIPAA-compliant retrospective evaluation of prospectively acquired data was approved by the local ethics committee. Written informed consent was obtained from all patients. All patients underwent mpMRI with T2W, DWI (ADC and b-2000 DWI), and DCE images with endorectal coil (ERC) using 3T MRI. Cases were consecutive patients who underwent mpMRI between 4/2012 and 6/2015 followed by radical prostatectomy (n=179). Controls were consecutive patients with imaging between 5/2013 and 5/2015 with no lesions detected on mpMRI and no history of a positive biopsy (n=92). Cases and controls were excluded for a hip prosthesis or for a missing sequence on mpMRI; cases were excluded if whole mount specimens were not available (total exclusions=69 cases, 38 controls). One control was excluded for converting to a positive biopsy during the course of the study. Thus, the total study population was 163 (n=110 cases; n=53 controls).
Among the 163 patients (n=110 cases; n=53 controls) the average age for cases and controls was 62.1 (range 41.4-83.7) and 61.9 (47.4-73.7) years old, respectively. The median PSA for cases and controls was 7.09 (1.7-84.6) and 5.50 (1.5-28.7) ng/mL, respectively. The median whole prostate volume for cases and controls was 41.2 (15.0-117.0) and 80.5 (31.0-160.0) mL, respectively.
Of the 163 patients included in this study, 34 were used in a previous multi-reader study with a different primary objective (publication pending).There was over a year and half between these studies and all patient information was anonymized between studies. Overlap was necessary as a finite number of patients with radical prostatectomy whole-mount specimens were available.
MR Imaging Protocol
The prostate mpMRI scans were acquired on a 3T scanner (Achieva 3.0T-TX, Philips Healthcare, Best, Netherlands) using an endorectal coil (BPX-30, Medrad, Pittsburgh, PA, USA) filled with 45ml fluorinert (3M, Maplewood, MN, USA) and the anterior half of a 32-channel cardiac SENSE coil (InVivo, Gainesville, FL, USA). [Table 1] contains the sequences and MRI acquisition parameters used in this study.
Study Design
Nine radiologists served as readers: 3 were highly experienced in prostate mpMRI (>2000 cases read last 2 years), 3 were moderately experienced (>500 cases read last 2 years), and 3 were inexperienced (<500 cases read last 2 years). Of the highly experienced readers, one was based in the United States (BT), one in the United Kingdom (TB), and one in Brazil (LB). Of the moderately experienced readers one was based in the United States (JM), one in Egypt (HS), and one in Singapore (YL). Of the inexperienced readers two were based in the United States (SB, RS) and one in Turkey (IK). In total, readers represented 6 nations and 8 institutions. All readers had experience with PIRADSv2 prior to starting this study.
To reach an achievable number of reads for each individual reader, the 163 patients were allocated to pairs of the 9 readers with 29 random cases evaluated by all 9 readers. Each pairwise combination of readers evaluated at least 4 patients assigned at random so each reader evaluated on average 59.4 patients (range 57-63). Patient allocation maintained a 2:1 ratio of cases to controls for each reader.
MRI Interpretation
Readers were blinded to all clinical and pathological outcomes. All patient identifying information was removed from images. RadiAnt DICOM Viewer ADDIN EN.CITE Medixant2015756(10)7567569MedixantRadiAnt DICOM Viewer2.2.9.107282015http://www.radiantviewer.com/(10) was used to view images. Readers were asked to detect up to three lesions on mpMRI that, in each readers judgment, would be included as part of a clinical report and score each lesion using PIRADSv2. Readers recorded the location of each detected lesion (TZ or PZ), presence of extra-prostatic extension (EPE), T2W score (1-5), DWI score (1-5), and DCE score (positive vs. negative) for each lesion according to PIRADSv2 criteria. Readers marked lesions on the dominant sequence (DWI for PZ and T2W for TZ) and saved a screen shot of all 4 sequences with the lesion marked [Figure 1]. All data were recorded in Microsoft Access.
Radiologic and Pathologic Correlation
All cases had whole-mount radical prostatectomy specimens produced with patient specific MRI based 3D-printed molds to optimize correlation between MRI and pathology specimen ADDIN EN.CITE ADDIN EN.CITE.DATA (11). Lesion locations and Gleason grades (Gl) on whole-mount specimens were annotated by a genitourinary pathologist. Correlation between reader pairs was made bas e d o n s c r e e n - s h o t s f r o m e a c h r e a d e r . C o r r e l a t i o n t o p a t h o l o g y w a s b a s e d o n p r o s t a t e l a n d m a r k s a n d l e s i o n m o r p h o l o g y . C l i n i c a l l y s i g n i f i c a n t ( C S ) d i s e a s e w a s d e f i n e d a s G l e"3 + 4 . F o r c o n t r o l c a s e s , 1 2 c o r e s y s t e m a t i c b i o p s y w a s u s e d t o v a l i d a t e M R I r e s u l t s .
Statistical Analysis
The overall PIRADS scores were determined using the scores given by a reader to a lesion (T2W, DWI, DCE, EPE), per the algorithm proposed in PIRADSv2 ADDIN EN.CITE Radiology2015380(1)380380017American College of RadiologyMR Prostate Imaging Reporting and Data System version 2.02015http://www.acr.org/Quality-Safety/Resources/PIRADS/12/2015(1). The cancer detection rate (CDR) was defined as the proportion of true positive lesions among all detected lesions.
To determine the efficacy of individual PIRADS scores, the CDR of all 9 readers were considered in a pooled analysis. For example, if one unique lesion in one patient was detected by all 9 readers and given PIRADS scores of 4, 4, 5, 5, 4, 3, 5, 4, 4 by each reader, then 9 total lesions would be considered. This analysis provides an assessment of PIRADSv2 performance across a broad spectrum of reader experience. As scores of the same lesion as well as multiple lesions in the same patient detected by multiple readers may be correlated, generalized estimation equations (GEE) with a logit link function under an independence working model assumption were used to obtain the robust variance estimates of CDRs, and the associated Wald test was used for inference.
The CDR was calculated for overall PIRADS scores and sequence-specific (T2W, DWI) PIRADS scores all zones, TZ, and PZ and compared to sequential scores and between zones. The CDRs for PIRADS 3+1 lesions in the PZ, or PIRADS=4 in the PZ by virtue of a DWI score of 3 and a positive DCE was compared to the CDRs of overall PIRADS=4 and PIRADS=3 by DWI=4 or 3, respectively, in the PZ.
To assess the performance of T2W and DWI in each zone, lesion based logistic regression prediction models were developed based on the individual sequence scores (T2W, DWI) and location (PZ, TZ) of a lesion. Similarly, to assess the incremental value of DCE positivity, the CDR for each PIRADS score on T2W and DWI with and without DCE positivity was estimated by a zonal-specific logistic regression model using separate T2W and DWI scores with and without DCE as a co-parameter. As T2W and DWI PIRADS scores exhibit a positive correlation with observed CDRs, both MRI sequences were treated as continuous linear predictors in the logistic models. For all models, variance estimates of the regression coefficients, odds-ratios, and Walds test were based on GEE under an independence working model assumption. Odds ratios (OR) were calculated as the ratio of odds of cancer of two consecutive scores. The fit of each predicted model was assessed by both graphical display of observed vs. model-based (predicted) CDRs and formally by testing for goodness of fit for the GEE model with binary responses ADDIN EN.CITE Barnhart1998768(12)76876817Barnhart, H. X.Williamson, J. M.Department of Biostatistics, Rollins School of Public Health of Emory University, Atlanta, Georgia 30322, USA. hbarnha@sph.emory.eduGoodness-of-fit tests for GEE modeling with binary responsesBiometricsBiometricsBiometricsBiometricsBiometricsBiometrics720-95421998/06/18Antidepressive Agents/therapeutic useComputer Simulation*Data Interpretation, StatisticalDepression/drug therapyDepressive Disorder/drug therapyDiabetes Mellitus, Type 1/epidemiology/physiopathologyDiabetic Retinopathy/epidemiologyFemaleHumans*Likelihood FunctionsLongitudinal StudiesMale*Models, StatisticalReproducibility of ResultsRisk FactorsWisconsin/epidemiology1998Jun0006-341X (Print)
0006-341x9629652NLMeng(12). All p-values correspond to two-sided tests, with a p-value <0.05 considered to represent a significant difference.
Results
Radiologic Lesion Characteristics
Among the 163 patients, a total of 654 lesions were detected by all readers comprising 336 unique lesions for an average of 2.06 lesions per patient. 234/654 lesions were in the TZ and 420/654 were in the PZ. For overall PIRADS scores of 1, 2, 3, 4, and 5 there were 3, 70, 115, 305, and 161 lesions, respectively. In the PZ there were 2, 41, 131, 171, and 75 lesions at PIRADS=1, 2, 3, 4, and 5, respectively. In the TZ there were 1, 28, 81, 56, and 68 lesions at PIRADS 1, 2, 3, 4, and 5, respectively. The PIRADS=1 lesions were excluded from further analysis based on small sample size (n=3).
Histopathologic Lesion Characteristics
Of the 110 cases, there were 268 total lesions on pathology or 2.4 per patient. There were 24, 151, 20, 63, 9, and 1lesions at Gl 3+3, 3+4, 4+3, 4+4, 4+5, and 5+4, respectively. 185/268 lesions were in the PZ with 18, 108, 17, 35, 6, and 1 at Gl 3+3, 3+4, 4+3, 4+4, 4+5, and 5+4 and 83/268 in the TZ with 6, 43, 3, 28, 3, and 0 at Gl 3+3, 3+4, 4+3, 4+4, 4+5, and 5+4, respectively. Of the 268 lesions, 29 demonstrated extra-prostatic extension on pathology.
Performance of PIRADSv2 Scores
The CDRs with standard error (SE) of the overall PIRADS scores are shown in [Table 2]. The CDR for PIRADS=2 was 24.3% for all lesions and 15.7% for CS disease. The CDR for PIRADS=3 was 40.0% for all lesions and 33.0% for CS disease. The CDR for PIRADS=4 was 78.7% for all lesions and 70.5% for CS disease. The CDR for PIRADS=5 was 91.3% for all lesions and 90.7% for CS disease. The increase in CDR between each increment in PIRADS score was statistically significant (p<0.05).
CDRs of Peripheral and Transition Zone
The CDRs with SE of the overall PIRADS score in each zone are shown in [Table 3]. For all lesions at PIRADS=2, the CDR in the PZ was 38.1% vs 3.6% in the TZ (p=0.001). For PIRADS=3, the CDR in the PZ was 51.1% vs 32.9% in the TZ (p=0.078). For PIRADS=4 the CDR was 81.9% in the PZ vs 66.1% in the TZ (p=0.039). For PIRADS=5 the CDR was 94.3% in the PZ vs 87.7% in the TZ (p=0.273). Only the difference at PIRADS=2 remained statistically significant for CS disease (23.8% in PZ vs. 3.6% in TZ; p=0.021).
Role of 3+1 Lesions
The relative CDR of each PIRADS scores fir PZ lesions rated PIRADS=3 on DWI and Positive on DCE (3+1 lesions) is shown in [Figure 2]. The CDR for 3+1 lesions was 67.8% for all lesions and 54.0% for CS lesions, significantly higher than overall PIRADS=3 lesions (40.0% vs 67.8%, p=0.02) and lower than pure PIRADS=4 lesions (67.8% vs 83.3%, p=0.002). It should be noted that only 4 lesions were called T2W=3 and DWI=5 in the TZ.
Validation of the Dominant Sequence
The predictive value of DWI vs T2W scores in the PZ for observed and modeled values is depicted in [Figure 3]. In the PZ, both DWI and T2W predicted models fit the data well (goodness-of-fit p=0.600 for DWI, 0.655 for T2W). However, DWI predicted a higher CDR for high-likelihood disease and a lower CDR for low-likelihood disease compared to T2W. The predicted CDRs for PIRADS=5 were 95.9% vs 92.7% and for PIRADS=2 were 35.3% vs 49.2% for DWI vs T2W. The OR for DWI vs T2W for increasing CDR for each incremental score was 3.49 vs 2.35 (p=0.008). Similar results were found for CS lesions with OR 4.11 vs 2.43 (p<0.001) for DWI vs T2W [Supplemental Figure 1]. The predicted CDR for DWI in the PZ was 35.3%, 65.6%, 86.9% and 95.9% for all lesions and 20.0%, 50.8%, 80.9% and 94.6% for CS lesions at PIRADS=2, 3, 4, 5, respectively. These results tend to validate the dominant sequence paradigm in the PZ.
For the TZ, the difference between the performance of DWI and T2W for all lesions was not statistically significant (OR DWI 4.79 vs OR T2W 3.77; p=0.494). However, the observed vs predicted CDR plot for the TZ shown in [Figure 4] trended for a good fit for T2W (goodness-of-fit p=0.521) which was not observed in DWI (goodness-of-fit p=0.102). Similar results were found for CS lesions (OR DWI 4.99 vs OR T2W 3.74, p=0.437) and displayed in [Supplemental Figure 2]. The predicted CDR for T2W in the TZ for all lesions was 12.4%, 34.7%, 66.7%, and 88.3% and for CS lesions was 11.7%, 33.2%, 65.0%, and 87.4% at PIRADS 2, 3,4, 5, respectively. Thus, the dominant sequence paradigm for the TZ was not convincingly demonstrated.
Benefit of DCE
The incremental benefit of DCE positivity at each PIRADS score was evaluated by the odds ratio obtained from the predicted model and by the predicted value for each DWI score. DCE positivity was an independent predictor for DWI in the PZ for all lesions and CS lesions. The OR was 2.0 (95% CI 1.08-3.70; p=0.027) for all lesions, indicating the odds of detecting cancer was two times higher when DCE was positive than when DCE was negative for any DWI score in the PZ. [Figure 5] displays the observed vs predicted CDR along with predicted CDR 95% confidence intervals for PIRADS DWI scores with and without DCE positivity. The predicted model fit the data well with the exception of DWI=5 where only 3 lesions were detected (goodness-of-fit p=0.747). For CS lesions in the PZ, the overall fit of the DWI-DCE predicted model was appropriate (goodness-of-fit p=0.356) with OR 2.45(95% CI 1.22-4.91; p=0.011) [Supplemental Figure 3].
For all lesions in the TZ, [Figure 6] displays the observed vs predicted CDR along with predicted CDR 95% confidence intervals for PIRADS T2W scores with and without DCE positivity. Due to the small number of lesions detected for the majority of T2W-DCE scores in the TZ, observed CDRs exhibit large variability which were not well accounted for by the predicted model (goodness-of-fit p=0.009). Nevertheless, the effect of DCE positivity for T2W in the TZ was significant and almost identical for all lesions (OR 2.8, 95% CI 1.09-7.16; p=0.032) and CS lesions (OR 2.9, 95% CI 1.14-7.39; p=0.026) [Supplemental Figure 4]. Thus, positive DCE parameters increase the likelihood of a positive biopsy for all cancers and for clinically significant cancers.
Discussion
This study suggests PIRADSv2 scores are predictive of cancer detection rates in the PZ and TZ. Our results demonstrate significant increases in the predictive value for each increment in PIRADS score for both all cancers and CS cancers (p<0.05 for all scores). This data affirms the use of the dominant sequence paradigm introduced in PIRADSv2 for PZ lesions where the dominant sequence is DWI. The dominant sequence paradigm could not be validated in the TZ, where the dominant sequence is T2W, with equivalent performance of DWI and T2W. This data also confirms the 3+1 strategy for employing DCE positivity to boost a PIRADS=3 score to PIRADS 4 in the PZ. However, these data also suggest that DCE provides a larger benefit across all PIRADS scores by increasing the cancer detection rate in the PZ at all scores. Again, this effect is seen in the TZ but is less convincing and a positive DCE does not provide as much benefit as it does in a PZ lesion
Each PIRADS score in the PZ and TZ result in approximately equivalent CDRs for CS disease except for PIRADS=2 (23.8% vs 3.6%; p=0.021). Our logistic regression models suggest that PIRADS DWI scores predict cancer more reliably in the PZ than T2W scores and T2W is at least equivalent to DWI in the TZ. These results provide evidence for the dominant sequence paradigm of PIRADSv2.
The role of DCE in prostate mpMRI has been the topic of much discussion ADDIN EN.CITE ADDIN EN.CITE.DATA (8, 9, 13, 14). The additional value of DCE is of particular interest. The use of DCE is increasingly controversial with several authors suggesting it be abandoned ADDIN EN.CITE ADDIN EN.CITE.DATA (13, 14). PIRADSv2 scoring criteria limits the role of DCE to the 3+1 strategy that permits a PIRADS 3 lesion to be increased to a PIRADS 4 lesion if the DCE is also positive. In the analysis of added benefit of DCE at each PIRADS score for each parameter in the PZ and TZ, DCE positivity improved the CDR of PIRADS=2, 3, and 4 lesions for the PZ (OR 2.00, p=0.027). While DCE trended towards a similar pattern in the TZ, no conclusive statistical significance was found with the relatively few lesions in the TZ in this cohort.
A potential weakness of PIRADSv2 is that the definitions of PIRADS 2, 3 and 4 lesions are vaguely defined and have caused some confusion among users. Rosenkrantz et al. ADDIN EN.CITE ADDIN EN.CITE.DATA (15) questioned how well these scores predicted cancer, the subjective terms used to classify each score, and the clinical implication (i.e., to biopsy or not to biopsy) for these scores. Mertan et al ADDIN EN.CITE ADDIN EN.CITE.DATA (16) found in a single reader, prospective study of PIRADSv2 that PIRADS=4 lesions highly underperformed compared to PIRADS=5 lesions for targeted biopsies (CDR 42.6% vs 78.1%, p=0.002). Our data suggest a positive DCE increases the likelihood of a positive biopsy and that the biopsy will show clinically significant disease at all PIRADS scores. Additionally, although PIRADS=5 and DCE negative lesions were a rare occurrence, (3 in the PZ and 15 in the TZ), the CDR for these lesions was lower than expected (66.7% DCE negative vs 92.3% DCE positive in the PZ and 46.7% vs 78.8% for the TZ respectively). This suggests DCE positivity may be useful for increasing the likelihood of disease for PIRADS=5 lesions. However, the role of DCE in lesion detection cannot be conclusively determined from these data. Prospective, controlled trials are necessary to elucidate the value of DCE.
This study revealed another interesting feature of PIRADSv2, the higher than expected CDR for PIRADS=2 lesions in the PZ. Conceptually, PIRADS=2 lesions do not require biopsy and readers in this study were instructed to detect lesions that would be included in a clinical report, which was p r e s u m e d t o b e P I R A D S e"3 . H o w e v e r , 7 0 P I R A D S = 2 l e s i o n s w e r e r e p o r t e d b y t h e r e a d e r s , 4 1 i n t h e P Z a n d 2 8 i n t h e T Z . W h i l e t h e C D R f o r a l l o f t h e s e l e s i o n s w a s l o w f o r C S d i s e a s e ( 1 5 . 7 % ) , t h e P Z P I R A D S = 2 l e s i o n s h a d a h i g h e r t h a n e x p e c t e d C D R o f 3 8 % f o r a l l lesions and 24% for CS lesions. While we did not study inter-reader variability, it is likely that there was some ambiguity in the mind of readers between PIRADS=2 and 3 lesions. In future revisions of PIRADS the subjectivity and resulting overlap between PIRADS 2 and 3 lesions should be considered when recommending biopsy based on scores.
There are several limitations to our study. Our study cohort relied on patients undergoing radical prostatectomy. This has the advantage of having certainty regarding the pathologic outcome of all lesions but tends to be a younger and higher risk population than the general at-risk population. To reduce detection bias, we included control patients without lesions detected on mpMRI and a negative systematic biopsy. Our study is also limited because it was performed at one institution. However, the imaging and interpretation protocols were entirely within PIRADSv2 guidelines. Furthermore, we included 9 readers who practice at 8 different institutions in 6 different countries to represent a broad population of readers with varying experience. Although inter-reader agreement was not a central theme of this study, we did account for inter-reader correlation in our GEE model, and the results are indicative of how readers across a broad spectrum of experience would score a detected lesion.
In conclusion, this study demonstrates that as the PIRADSv2 scores incrementally increase, the cancer detection rate also increases both for all prostate cancers and for clinically significant prostate cancers. This study affirms the value of the dominant sequence paradigm put forth in PIRADSv2 especially in the PZ and clarifies that DCE MRI, whose role in PIRADSv2 is controversial, adds significant benefit to PIRADS 3 and 4 lesions in the PZ. Rather than suggesting DCE should be eliminated, these data suggest DCE should be expanded to other PIRADS scores to stratify risk more accurately. In short, PIRADSv2 is a practical, reproducible and accurate method of assessing the risk of cancer in multiparametric MRI of the prostate.
References
ADDIN EN.REFLIST 1. Radiology ACo. MR Prostate Imaging Reporting and Data System version 2.0. 2015.
2. Vache T, Bratan F, Mege-Lechevallier F, Roche S, Rabilloud M, Rouviere O. Characterization of prostate lesions as benign or malignant at multiparametric MR imaging: comparison of three scoring systems in patients treated with radical prostatectomy. Radiology. 2014;272(2):446-55.
3. Delongchamps NB, Rouanne M, Flam T, et al. Multiparametric magnetic resonance imaging for the detection and localization of prostate cancer: combination of T2-weighted, dynamic contrast-enhanced and diffusion-weighted imaging. BJU international. 2011;107(9):1411-8.
4. Akin O, Sala E, Moskowitz CS, et al. Transition zone prostate cancers: features, detection, localization, and staging at endorectal MR imaging. Radiology. 2006;239(3):784-92.
5. Hoeks CM, Hambrock T, Yakar D, et al. Transition zone prostate cancer: detection and localization with 3-T multiparametric MR imaging. Radiology. 2013;266(1):207-17.
6. Turkbey B, Shah VP, Pang Y, et al. Is apparent diffusion coefficient associated with clinical risk scores for prostate cancers that are visible on 3-T MR images? Radiology. 2011;258(2):488-95.
7. Rosenkrantz AB, Mannelli L, Kong X, et al. Prostate cancer: utility of fusion of T2-weighted and high b-value diffusion-weighted images for peripheral zone tumor detection and localization. Journal of magnetic resonance imaging : JMRI. 2011;34(1):95-100.
8. Iwazawa J, Mitani T, Sassa S, Ohue S. Prostate cancer detection with MRI: Is dynamic contrast-enhanced imaging necessary in addition to diffusion-weighted imaging? Diagnostic and Interventional Radiology. 2011;17(3):243-8.
9. Junker D, Quentin M, Nagele U, et al. Evaluation of the PI-RADS scoring system for mpMRI of the prostate: a whole-mount step-section analysis. World journal of urology. 2015;33(7):1023-30.
10. Medixant. RadiAnt DICOM Viewer. 2.2.9.10728 ed. HYPERLINK "http://www.radiantviewer.com/2015" http://www.radiantviewer.com/2015.
11. Shah V, Pohida T, Turkbey B, et al. A method for correlating in vivo prostate magnetic resonance imaging and histopathology using individualized magnetic resonance-based molds. The Review of scientific instruments. 2009;80(10):104301.
12. Barnhart HX, Williamson JM. Goodness-of-fit tests for GEE modeling with binary responses. Biometrics. 1998;54(2):720-9.
13. Rud E, Baco E. Re: Jeffrey C. Weinreb, Jelle O. Barentsz, Peter L. Choyke, et al. PI-RADS Prostate Imaging - Reporting and Data System: 2015, Version 2. Eur Urol 2016;69:16-40: Is Contrast-enhanced Magnetic Resonance Imaging Really Necessary When Searching for Prostate Cancer? European urology. 2016.
14. Barentsz JO, Choyke PL, Cornud F, et al. Reply to Erik Rud and Eduard Baco's Letter to the Editor re: Re: Jeffrey C. Weinreb, Jelle O. Barentsz, Peter L. Choyke, et al. PI-RADS Prostate Imaging - Reporting and Data System: 2015, Version 2. Eur Urol 2016;69:16-40. European urology. 2016.
15. Rosenkrantz AB, Oto A, Turkbey B, Westphalen AC. Prostate Imaging Reporting and Data System (PI-RADS), Version 2: A Critical Look. AJR American journal of roentgenology. 2016:1-5.
16. Mertan FV, Greer MD, Shih JH, et al. Prospective Evaluation of the Prostate Imaging Reporting and Data System version 2 (PI-RADSv2) for Prostate Cancer Detection. The Journal of urology. 2016.
Tables
Table 1
Table 1: Multiparametric MR Imaging Sequence Parameters at 3TParameterT2 WeightedDWI*High b-Value DWIDCE MR ImagingField of view (mm)140 140140 140140 140262 262Acquisition Matrix304 234112 10976 78188 96Repetition time (msec)4434498669873.7Echo time (msec)12054522.3Flip angle (degrees)9090908.5Section thickness (mm), no gaps3333Image reconstruction matrix (pixels)512 512256 256256 256256 256Reconstruction voxel imaging resolution (mm/pixel)0.27 0.27 3.000.55 0.55 2.730.55 0.55 2.731.02 1.02 3.00Time for acquisition (min:sec)2:484:543:505:16* For ADC map calculation. Fi v e e v e n l y - s p a c e d b v a l u e s ( 0 - 7 5 0 s e c / m m 2 ) w e r e u s e d b = 2 0 0 0 s e c / m m 2
T a b l e 2
C a n c e r d e t e c t i o n r a t e s w i t h s t a n d a r d e r r o r f o r o v e r a l l P I R A D S s c o r e s f o r a l l l e s i o n s a n d c l i n i c a l l y s i g n i f i c a n t l e s i o n s P I R A D S e"3 + 3 p - v a l u e * e"3 + 4 p - v a l u e * 2 2 4 . 3 % ( 6 . 6)15.7% (5.4)340.0% (6.2)0.02533.0% (5.9)0.006478.7% (3.5)<0.00170.5% (4.6)<0.001591.3% (2.6)0.00190.7% (2.8)<0.001*p-value between consecutive PIRADS scoresPZ = peripheral zone, TZ = transition zone
Table 3
Cancer detection rate with standard error for all lesions and clinically significant lesions for overall PIRADS scores between the PZ and TZPIRADS e"3 + 3 e"3 + 4 P Z T Z p - v a l u e * P Z T Z p - v a l u e * 2 3 8 . 1 % ( 9 . 4 ) 3 . 6 % ( 3 . 6 ) 0 . 0 0 1 2 3 . 8 % ( 8 . 0 ) 3 . 6 % ( 3 . 6 ) 0 . 0 2 1 3 5 1 . 1 % ( 8 . 9 ) 3 2 . 9 % ( 7 . 1 ) 0 . 0 7 8 3 7 . 8 % ( 9 . 1 ) 3 0 . 0 % ( 7 . 1 ) 0 . 4 7 8 4 8 1 . 9 % ( 3 . 6 ) 6 6 . 1 % ( 7 . 4 ) 0 . 0 3 9 7 2 . 0 % ( 5 . 2 ) 6 4 . 5 % ( 7 . 4 ) 0 . 3 7 8 5 9 4 . 3 % ( 2 . 6 ) 8 7 . 7 % ( 5 . 1 ) 0.27393.2% (3.3)87.7% (5.1)0.388*p-value between PZ and TZ
PZ = peripheral zone, TZ = transition zone
Figures
Figure 1
Screenshots for 9 readers. A 66 year old man with a PSA of 8.3ng/ml who underwent ERC 3T mpMRI with T2W, ADC, b-2000, and DCE imaging followed by prostatectomy. Nine readers were asked to detect all lesions that would be included in a clinical report and score those with PIRADSv2. Shown are the T2W screen shots of all 9 readers of an anterior mid transition zone lesion that was Gleason=3+4 on prostatectomy, with histopathology demonstrated on right.
Figure 2
Cancer detection rate (CDR) with standard error for each PIRADSv2 score. Each score demonstrated an added benefit over the previous score (p<0.05). DWI=3 and DCE positive lesions in the PZ (3+1) represent a distinct risk population from all other PIRADS=4 and PIRADS=3 lesions. (* p<0.05)
Figure 3
Validation of the dominant parameter in the PZ. The PIRADS scores for DWI and T2W in the PZ are shown with corresponding odds ratios (OR) and p-value for goodness of fit. PIRADS DWI scores demonstrated a higher predictive value for high likelihood scores (PIRADS=4, 5) and lower predictive value for low likelihood disease (PIRADS=2) vs PIRADS T2W scores (OR=3.49 vs 2.35; p=0.008)
Figure 4
Validation of the dominant parameter in the TZ. The PIRADS scores for DWI and T2W in the TZ are shown with corresponding odds ratios (OR) and p-value for goodness of fit. No clear separation in incremental prediction value of PIRADS scores was seen (OR 3.77 vs 4.79; p=0.494). However, the observed vs predicted plot trended for a better fit for T2W (goodness-of-fit p=0.521) over DWI (goodness-of-fit p=0.102).
Figure 5
Incremental value of DCE in the PZ to DWI scoring for all lesions. Displayed are the observed CDRs for each PIRADS DWI score in the PZ with DCE negativity or positivity and PIRADS DWI score as independent variables the model-based prediction with 95% confidence interval bars. Total counts of lesions detected at a given PIRADS score are displayed in red below each score. DCE positivity demonstrated an OR of 2.0 (95% CI 1.08-3.70; p=0.027) for cancer positivity.
Figure 6
Incremental value of DCE in the TZ to T2W scoring for all lesions. Displayed are the observed CDRs for each PIRADS T2W score in the TZ with DCE negativity or positivity and PIRADS T2W scores as independent variables in the model-based prediction with 95% confidence interval bars. Total counts of lesions detected at a given PIRADS score are displayed in red below each score. DCE positivity demonstrated an OR of 2.8 (95% CI 1.09-7.16; p=0.032) for cancer positivity. However, the model is limited by fit to the observed values (p=0.009).
Supplemental Figures
Supplemental Figure 1
Validation of the dominant parameter in the PZ for clinically significant (CS) lesions (Gleason e" 3 + 4 ) . T h e P I R A D S s c o r e s f o r D W I a n d T 2 W i n t h e P Z a r e s h o w n w i t h c o r r e s p o n d i n g o d d s r a t i o s ( O R ) a n d p - v a l u e f o r g o o d n e s s o f f i t . P I R A D S D W I s c o r e s d e m o n s t r a t e d a h i g h e r p r e d i c t i v e v a l u e f o r h i g h l i k e l i h o o d s c o r e s a n d l o w e r p r e d i c t i v e v a l u e f o r l o w l i k e l i h o o d d i s e a s e v s P I R A D S T 2 W s c o r e s ( O R = 4 . 1 1 v s 2 . 4 3 ; p <