Genitourinary Imaging Original Research

Transcription

1 Genitourinary Imaging Original Research Kitajima et al. MRI of Local Recurrence After Prostatectomy Genitourinary Imaging Original Research Kazuhiro Kitajima 1,2 Robert P. Hartman 1 Adam T. Froemming 1 Clinton E. Hagen 3 Naoki Takahashi 1 Akira Kawashima 1 Kitajima K, Hartman RP, Froemming AT, Hagen CE, Takahashi N, Kawashima A Keywords: DWI, dynamic contrast-enhanced imaging, local recurrence, prostate cancer, prostatectomy DOI: /AJR Received December 9, 2014; accepted after revision February 9, Department of Radiology, Mayo Clinic, 200 First St SW, Rochester, MN Address correspondence to A. Kawashima (kawashima.akira@mayo.edu). 2 Present address: Department of Radiology, Kobe University Graduate School of Medicine, Kobe, Japan. 3 Department of Biomedical Statistics and Informatics, Mayo Clinic, Rochester, MN. AJR 2015; 205: X/15/ American Roentgen Ray Society Detection of Local Recurrence of Prostate Cancer After Radical Prostatectomy Using Endorectal Coil MRI at 3 T: Addition of DWI and Dynamic Contrast Enhancement to T2-Weighted MRI OBJECTIVE. The purpose of this study was to assess the utility of DWI and dynamic contrast enhancement (DCE) in MRI at 3 T with an endorectal coil in identifying local prostate cancer recurrence after radical prostatectomy. MATERIALS AND METHODS. Eighty men underwent MRI for suspected local recurrence. The reference standards were histopathologic result, decrease in prostate-specific antigen level after salvage radiation therapy, and follow-up findings. Using a 5-point scoring system, two reviewers independently interpreted T2-weighted images alone (protocol A), a combination of T2-weighted and DW images (protocol B), a combination of T2-weighted and DCE images (protocol C), and a combination of T2-weighted, DW, and DCE images (protocol D). ROC analysis was used to compare the four protocols. RESULTS. Local recurrence was found in 57 of the 80 patients (71.3%). The ROC AUCs for protocols A, B, C, and D were 0.71, 0.72, 0.90, and 0.89 for reader 1 and 0.65, 0.62, 0.84, and 0.83 for reader 2. Protocols C and D had statistically better performance than protocols A and B for both readers (p < 0.001). For local recurrence lesions with a long-axis diameter less than 10 mm visualized with protocol C, protocol B had detection rates of % and for lesions measuring 10 mm or greater, %. The rates of detection of local recurrence with protocol C or D were % for prostate-specific antigen levels less than 0.4 ng/ml, % for ng/ml, and % for greater than 1.0 ng/ml. CONCLUSION. Addition of DCE to T2-weighted imaging in 3-T MRI with an endorectal coil improves the accuracy of detection of local cancer recurrence after radical prostatectomy. The addition of DWI is of limited incremental value for detection, especially of small lesions. M any men with prostate cancer, which is the most common malignant tumor and the second leading cause of cancer death in the United States, are treated with definitive therapy, including radical prostatectomy (RP) and radiation therapy. The first sign of recurrence is often the return of measurable prostate-specific antigen (PSA). Biochemical recurrence after RP, which is defined as a PSA level of 0.2 ng/ml or greater followed by another increased value [1], has been reported to occur in 20 40% of cases [2, 3] and is associated with increased risk of disease progression or death. It is important to determine whether local recurrence, distant metastatic disease, or a combination of the two is present because this determination affects therapy, including salvage and systemic treatment options [4]. The diagnosis of local recurrence is often difficult and poses a dilemma. Transrectal ultrasound guided biopsy of the prostatic fossa is invasive and not sensitive at low PSA levels. The sensitivity of prostatic fossa biopsy has been reported to be 21 26% for PSA levels of 1 ng/ml or less and 49 77% for PSA levels greater than 1 ng/ml, and repeated biopsies are frequently required [5, 6]. An accurate and noninvasive imaging method of detecting local recurrence as early as possible is therefore desired. The addition of dynamic contrast enhancement (DCE) to T2-weighted MRI with an endorectal coil has been found useful for visualizing local tumor recurrence after RP. The images show excellent contrast between hyperenhancing recurrent tumor and hypoenhancing noncancerous soft tissue [7 12]. DWI has also been found useful in detecting prostate carcinoma before treatment and recurrence after radiation therapy [13, 14]. However, evaluation of DWI for depicting recurrent tumor in the prostatectomy bed has been limited [15 18]. Among four reports, one study was conducted AJR:205, October

2 Kitajima et al Sensitivity Specificity ROC Curve (Area) Protocol A (0.7067) Protocol B (0.7243) Protocol C (0.8966) Protocol D (0.8856) A with a combination of DWI and DCE [15] and another with a combination of DWI, DCE, and spectroscopic MRI [16] without data on the individual components separately. To our knowledge, only two studies [17, 18] have assessed the performance of DWI separately. Moreover, among previous reports showing the diagnostic accuracy of MRI for detecting local tumor recurrence after RP, 1.5-T MRI systems were used in five studies [8 12] and 3-T MRI systems in only four [15 18]. Finally, in the previously reported studies [8 12, 15 18], patients who had undergone salvage treatment after RP were excluded. Therefore, the utility of MRI in evaluating patients with suspected tumor recurrence in the prostatectomy bed after previous second-line treatment has not been defined, to our knowledge. The purposes of our study were to evaluate the utility of 3-T endorectal coil multiparametric MRI for identifying suspected local tumor recurrence in the prostatectomy bed in both patients who have undergone and in those who have not undergone previous second-line treatment and to define the incremental and combined values of the addition of DCE and DWI to T2-weighted MRI. Sensitivity Specificity ROC Curve (Area) Protocol A (0.6472) Protocol B (0.6178) Protocol C (0.8383) Protocol D (0.8345) B Fig. 1 ROC curves of four image protocols for detection of localized recurrent prostate carcinoma in 80 prostatectomy patients. With addition of dynamic contrastenhanced MRI (DCE MRI), AUCs of combining T2-weighted and DCE images (protocol C) and combining T2-weighted, DW, and DCE images (protocol D) are significantly larger for both readers than for T2-weighted imaging alone (protocol A) and combined T2-weighted imaging and DWI (protocol B). Addition of DWI to T2-weighted and DCE imaging did not significantly change AUCs for either reader. A, Graph shows that for reader 1, protocols C (p = ) and D (p = ) outperformed protocol A (p = ) and also outperformed protocol B (p = ). B, Graph shows that for reader 2, protocols C (p = ) and D (p = ) outperformed protocol A (p < ) and also outperformed protocol B (p = ). Materials and Methods Patients This retrospective study was conducted with the approval of the institutional review board; it was exempt from the requirement for informed patient consent. A total of 280 consecutively registered patients with suspected local recurrence after RP who underwent 3-T endorectal coil multiparametric MRI, which included T2-weighted imaging, DCE, and DWI, between August 2010 and April 2012 were enrolled in this study. The standard of reference for defining local recurrence included positive result of transrectal ultrasound guided biopsy of the prostatectomy bed, greater than 50% reduction in serum PSA level after salvage therapy, and increase in PSA level and MRI finding of increase in size of a suspicious lesion during a 6-month follow-up period [10, 16, 17]. A patient was considered free of local recurrence only when benign histopathologic findings were obtained at biopsy and PSA level returned to undetectable without any form of therapy for at least three consecutive measurements over at least 1 year of follow-up and findings were normal at follow-up MRI [8, 10, 12]. Eighty-three of the 280 patients met at least one of the three criteria for defining presence or absence of recurrence. Three of the 83 were excluded because of susceptibility artifact from hip arthroplasty. The other 80 patients constituted the final study group and were also stratified into two groups depending on whether they had undergone second-line treatment after RP but before MRI. MRI Technique All subjects were examined with a 3-T MRI system (Discovery-MR750, GE Healthcare) with a combination of an eight-channel phased-array coil and an endorectal coil, which was insufflated with 50 ml of 60% weight/volume barium solution. Our standard MRI protocol includes axial, coronal, and sagittal fast spin-echo T2-weighted imaging, axial DWI, and axial DCE MRI with a small FOV covering the surgical bed. T2-weighted imaging was performed with the following parameters: TR/TE, /98 110; slice thickness, 2.5 mm; gap, 0.5 mm; FOV, cm 2 ; matrix, ; number of signals acquired, 2 or 3; flip angle, 90. Spin-echo echo-planar DWI was performed with the following parameters: TR/TE, 4000/64 66; slice thickness, 5 mm; gap, 1 mm; FOV, 22 cm; matrix, ; number of signals acquired, 6; flip angle, 90 ; b values, 0 and 1000 s/mm 2 for 29 patients imaged from August 2010 to February AJR:205, October 2015

3 MRI of Local Recurrence After Prostatectomy or 0, 600, and 1000 s/mm 2 for 51 patients imaged from March 2011 to April For DCE MRI acquisition, 3D fast spoiled gradient-echo images (liver acquisition with volume acquisition) were acquired before, during, and after IV administration of a total of 0.1 mmol/kg of gadodiamide (Omniscan, GE Healthcare) at a rate of 2 ml/s through a power injector (Spectris, Medrad) and followed by a saline flush (15 ml at 2 ml/s). Three-dimensional fast-spoiled gradient-echo T1-weighted fat-suppressed sequences were performed with the following parameters: TR/TE, 5.2/2.6; slice thickness, 2.6 mm; overlap, 1.3 mm; FOV, 18 cm 2 ; matrix, ; number of signals acquired, 0.72; flip angle, 12. Multiphase DCE MRI sequences were performed every 30 seconds before and after IV contrast administration for 4 5 minutes. A C B Fig year-old man with increased prostate-specific antigen level of 6.2 ng/ml 7 months after failed radical prostatectomy. Reader 1 assigned scores of 3 for protocol A, 2 for protocol B, and 4 for both C and D. Reader 2 assigned scores of 1 for protocol A, 1 for protocol B, and 4 for C and D. Histopathologic findings obtained at transrectal ultrasound guided biopsy revealed prostate cancer (Gleason 3 + 3). Transurethral catheter is evident in all images. A, Axial T2-weighted MR image shows slightly hyperintense area (arrow) in right anterolateral bladder neck at vesicourethral anastomosis. B, Apparent diffusion coefficient map shows no apparent restricted diffusion. C, Dynamic contrast-enhanced MR image shows 6 11 mm avidly enhancing nodule (arrow) at right anterolateral bladder neck. Finding suggests localized recurrence of prostate cancer. Image Analysis At a PACS workstation (Centricity RA1000, GE Healthcare), two experienced genitourinary radiologists with 22 and 11 years of prostate MRI experience retrospectively and independently assessed all MR images obtained with four imaging protocols. They were aware that the patients presented with post-rp biochemical recurrence but were blinded to the other clinical and histopathologic information and other imaging findings. The protocols were as follows: protocol A consisted of T2-weighted imaging data alone; protocol B, data from T2-weighted imaging and DWI and apparent diffusion coefficient (ADC) mapping; protocol C, data from T2-weighted imaging and DCE MRI; protocol D, data from the combination of T2-weighted imaging, DWI and ADC mapping, and DCE MRI. Images in each of the four protocols were read at random by the two reviewers after a minimum interval of 2 weeks between reading sessions to minimize recall bias due to reading-order effects. For each imaging protocol for each patient, a confidence score reflecting the perceived likelihood and location of local recurrent tumors was assigned according to the following 5-point scale: 5, definitively present; 4, probably present; 3, equivocal; 2, probably absent; and 1, definitively absent. The following diagnostic criteria were applied to the existing literature data [7 12, 17] and our clinical experience. When a T2-weighted image was reviewed, a mass or soft-tissue area showing slightly high signal intensity relative to muscle was considered a positive finding (confidence score, 5 or 4). In review of DW images and ADC maps, the presence of a lesion with high focal signal intensity on the DW image and low signal intensity on the ADC map relative to background was considered a positive finding (confidence score, 5 or 4). In a review of DCE MR images, a hyperenhancing lesion on early phase images (< 90 seconds after gadolinium injection) with rapid washout or plateau on delayed images was considered a positive finding (confidence score, 5 or 4). In protocols B and C, two positive interpretations of images in two sequences was defined as a confidence score of 5, and one positive interpretation in two sequences was defined as a score of 4 or 3. In protocol D, three positive interpretations in three sequences was defined as a con- AJR:205, October

4 Kitajima et al. fidence score of 5; two positive interpretations in three sequences, 5 or 4; and one positive interpretation, 4 or 3. Both reviewers recorded the long and short dimensions of suspicisous lesions on MR images obtained with each imaging protocol. A B Fig year-old man with increased prostatespecific antigen level of 3.9 ng/ml 4.2 years after failed radical prostatectomy followed by salvage external beam radiation therapy and androgen deprivation therapy. Both readers assigned score of 5 with all four protocols. Histopathologic examination after transrectal ultrasound guided biopsy revealed prostate cancer (Gleason 5 + 5). A, Axial T2-weighted image shows lobulated hyperintense mass (arrow) measuring mm within left anterolateral bladder neck. B, Apparent diffusion coefficient map shows lesion (arrow) has low signal intensity (apparent diffusion coefficient, mm 2 /s). C, Dynamic contrastenhanced MR image shows fast and avid enhancement of lesion (arrow). C Statistical Analysis Interobserver agreement was determined per patient. Interobserver agreement was considered slight when less than 0.21, fair when ranging from 0.21 to 0.40, moderate when ranging from 0.41 to TABLE 1: Patient and Original Tumor Characteristics (n = 80) Characteristic Value Age (y) Mean 67.2 Range Prostate-specific antigen level (ng/ml) at MRI Mean 1.17 Range Median 0.43 Interval between RP and multiparametric MRI (mo) Mean 75.4 Range Treatment before MRI RP only 49 (61.3) RP plus salvage EBRT 11 (13.8) RP + ADT 10 (12.5) RP + salvage EBRT + ADT 10 (12.5) Pathologic stage pt2n0 45 (56.3) pt3an0 17 (21.2) pt3bn0 12 (15.0) Any TpN1 6 (7.5) No. with positive surgical margin 43 (53.8) Pathologic Gleason score at RP (20.0) 7 (3 + 4) 31 (38.8) 7 (4 + 3) 18 (22.5) (18.8) Note Unless otherwise indicated, values are number of patients with percentage in parentheses. Percentages may not total 100 owing to rounding. RP = radical prostatectomy, EBRT = external beam radiation therapy, ADT = androgen deprivation therapy. 0.60, substantial when ranging from 0.61 to 0.80, and almost perfect when greater than 0.80 [19]. To estimate the utility of each protocol in the diagnosis of local recurrence of prostate cancer, ROC analysis was conducted. To test whether the ROC AUCs were different, correlation of the testing methods was accounted for in the analysis [20]. The McNemar test was used to assess differences in sensitivity, specificity, and accuracy between two protocols. To calculate the sensitivity, specificity, and accuracy of each protocol, confidence scores of 4 and 5 were considered positive. Tests for differences in sensitivity, specificity, and accuracy of each protocol between subgroups of patients without and with previous second-line treatment were 810 AJR:205, October 2015

5 MRI of Local Recurrence After Prostatectomy TABLE 2: Diagnostic Performance of Four Image Protocols of 3-T Endorectal Coil MRI for Detection of Local Recurrence After Radical Prostatectomy (RP) All patients (n = 80) Reader 1 Protocol Sensitivity (%) Specificity (%) Accuracy (%) AUC Optimal Cutoff Point A 59.6 [34/57] ( ) 87.0 [20/23] ( ) 67.5 [55/80] ( ) 0.71 ( ) B 45.6 [26/57] ( ) 91.3 [21/23] ( ) 59.3 [47/80] ( ) 0.72 ( ) C 82.5 [47/57] ( ) 87.0 [20/23] ( ) 83.8 [67/80] ( ) 0.90 ( ) D 82.5 [47/57] ( ) 87.0 [20/23] ( ) 83.8 [67/80] ( ) 0.89 ( ) Comparisons C and D > A; A, C, D > B C and D > A; C and D > B C and D > A; C and D > B Reader 2 A 54.4 [31/57] ( ) 69.6 [16/23] ( ) 58.8 [47/80] ( ) 0.65 ( ) B 42.1 [24/57] ( ) 78.3 [18/23] ( ) 53.1 [42/80] ( ) 0.62 ( ) C 73.7 [42/57] ( ) 82.6 [19/23] ( ) 76.3 [61/80] ( ) 0.84 ( ) D 71.9 [41/57] ( ) 87.0 [20/23] ( ) 76.3 [61/80] ( ) 0.83 ( ) Comparisons C and D > A; A, C, D > B C and D > A; C and D > B C and D > A; C and D > B Patients with no previous second-line therapy between RP and MRI (n = 49) Reader 1 A 60.0 [21/35] ( ) 92.9 [13/14] ( ) 69.4 [34/49] ( ) 0.75 ( ) B 44.4 [15/35] ( ) 92.9 [13/14] ( ) 57.1 [28/49] ( ) 0.76 ( ) C 77.1 [27/35] ( ) 92.9 [13/14] ( ) 81.6 [40/49] ( ) 0.91 ( ) D 77.1 [27/35] ( ) 92.9 [13/14] ( ) 81.6 [40/49] ( ) 0.88 ( ) Comparisons C and D > A; A, C, D > B C and D > A; A, C, D > B C and D > A; C and D > B Reader 2 A 55.6 [19/35] ( ) 78.6 [11/14] ( ) 61.2 [30/49] ( ) 0.69 ( ) B 41.7 [14/35] ( ) 78.6 [11/14] ( ) 52.0 [25/49] ( ) 0.62 ( ) C 74.3 [26/35] ( ) 85.7 [12/14] ( ) 77.6 [38/49] ( ) 0.83 ( ) D 71.4 [25/35] ( ) 92.9 [13/14] ( ) 77.6 [38/49] ( ) 0.84 ( ) Comparisons C > A; A, C, D > B C and D > A; C and D > B C and D > A; C and D > B Patients with previous second-line therapy between RP and MRI (n = 31) Reader 1 A 59.1 [13/22] ( ) 77.8 [7/9] ( ) 64.5 [20/31] ( ) 0.65 ( ) B 50.0 [11/22] ( ) 88.9 [8/9] ( ) 61.3 [19/31] ( ) 0.67 ( ) C 90.9 [20/22] ( ) 77.8 [7/9] ( ) 87.1 [27/31] ( ) 0.91 ( ) D 90.9 [20/22] ( ) 77.8 [7/9] ( ) 87.1 [27/31] ( ) 0.91 ( ) Comparisons C and D > A; C and D > B C and D > A; C and D > B C and D > A; C and D > B Reader 2 A 54.5 [12/22] ( ) 55.6 [5/9] ( ) 54.8 [17/31] ( ) 0.60 ( ) B 45.5 [10/22] ( ) 77.8 [7/9] ( ) 54.8 [17/31] ( ) 0.61 ( ) C 72.7 [16/22] ( ) 77.8 [7/9] ( ) 74.2 [23/31] ( ) 0.85 ( ) D 72.7 [16/22] ( ) 77.8 [7/9] ( ) 74.2 [23/31] ( ) 0.83 ( ) Comparisons C and D > A; C and D > B C and D > A; C and D > B C and D > A; C and D > B Note Values in square brackets are numbers of patients. Values in parentheses are 95% CI. Scores of 4 and 5 were considered positive in calculation of sensitivity, specificity, and accuracy of each protocol. In comparisons, > symbol indicates statistically significant difference (p < 0.05) between protocol or protocols to left of symbol and protocol to right of symbol. Youden Index AJR:205, October

6 Kitajima et al. conducted with the Fisher exact test. In addition to ROC analysis, the optimal cutoff point was determined by use of the maximum Youden index for each protocol. Descriptive rates of detection of local recurrence lesions with MRI were stratified according to serum PSA level (< 0.4 ng/ml, ng/ml, > 1.0 ng/ml) and maximum lesion length on MR images ( 5 mm, 6 9 mm, mm, 20 mm). Values of p < 0.05 were considered to indicate a statistically significant difference in all analyses. Statistical analysis was performed with SAS software (version 9.3, SAS Institute). Results The characteristics of the 80 patients and their original tumors are presented in Table 1. TABLE 3: Rate of Detection of Local Tumor Recurrence per Patient Stratified by Serum Prostate-Specific Antigen Levels for MRI Protocols Prostate-Specific Antigen Level (ng/ml) No. of Patients With Recurrence All Patients In 57 of the 80 patients (71.3%), local recurrence was confirmed with histopathologic findings obtained at transrectal ultrasound guided biopsy (n = 31), subsequent reduction in PSA level after local salvage therapy (n = 20), and MRI and PSA followup findings (n = 6). Interobserver agreement was substantial for all protocols: protocol A, κ = 0.64; protocol B, κ = 0.68; protocol C, κ = 0.74; and protocol D, κ = The per-patient sensitivity, specificity, accuracy, and ROC AUC for the four protocols are shown in Table 2 and Figure 1. The addition of DCE to T2-weighted imaging (protocol C) and to T2-weighted imaging and DWI (protocol D) showed statistically significant larger AUCs than did protocols A (T2-weighted imaging only) and B (T2-weighted imaging and DWI) for both readers (p < 0.001). However, protocols B and D with the addition of DWI did not exhibit a significant difference in AUC from the AUCs of protocols A and C for either reader. For each reader, protocols C and D had significantly higher sensitivity and accuracy than protocols A and B (p < 0.001). Protocol A was more sensitive than protocol B for both readers (p < 0.05). No significant difference in specificities was present among the protocols for either reader. Thirty-three patients had a serum PSA level less than 0.4 ng/ml; 19, ng/ml; and 28, greater than 1.0 ng/ml. Local recurrence was documented in 17 of the 33 patients (51.5%), 15 of the 19 (78.9%), and 25 of the 28 (89.3%). With protocols C and D, reader 1 correctly detected local recurrences with confidence scores of 5 or 4 in 47 of the 57 patients (82.5%) with local recurrence. With protocol C, the rates of detection of recurrent tumor with MRI were 82.4% (14/17) for PSA levels less than 0.4 ng/ml, 73.3% (11/15) for PSA levels of ng/ml, and 88.0% (22/25) for PSA levels greater than 1 ng/ml (Table 3). Although reader 2 correctly found recurrent lesions in 42 of the 57 patients (73.7%) with documented recurrent disease with scores of 5 or 4 using protocol C, the rates of detection of tumor recurrence with MRI were 76.5% (13/17) for a PSA level less than 0.4 ng/ml, 60.0% (9/15) for a PSA level of ng/ml, and 80.0% (20/25) for a PSA level greater than 1 ng/ml (Table 3). MRI (protocols C and D) revealed multiple separate lesions in the prostatectomy bed No. of Patients With True-Positive MRI Result Protocol A Protocol B Protocol C Protocol D Reader 1 All 80 patients < (52.9) 8 (47.1) 14 (82.4) 14 (82.4) (66.7) 6 (40.0) 11 (73.3) 11 (73.3) > (60.0) 12 (48.0) 22 (88.0) 22 (88.0) 49 Patients without second-line therapy < (53.8) 5 (38.5) 10 (76.9) 10 (76.9) (70.0) 4 (40.0) 7 (70.0) 7 (70.0) > (66.7) 6 (50.0) 10 (83.3) 10 (83.3) 31 Patients with previous second-line therapy < (50.0) 3 (75.0) 4 (100) 4 (100) (60.0) 2 (40.0) 4 (80.0) 4 (80.0) > (53.8) 6 (46.2) 12 (92.3) 12 (92.3) Reader 2 All 80 patients < (41.2) 6 (35.3) 13 (76.5) 13 (76.5) (53.3) 7 (46.7) 9 (60.0) 8 (53.3) > (64.0) 11 (44.0) 20 (80.0) 20 (80.0) 49 Patients without second-line therapy < (38.5) 3 (23.1) 10 (76.9) 10 (76.9) (60.0) 5 (50.0) 7 (70.0) 6 (60.0) > (66.7) 6 (50.0) 9 (75.0) 9 (75.0) 31 Patients with previous second-line therapy < (50.0) 3 (75.0) 3 (75.0) 3 (75.0) (40.0) 2 (40.0) 2 (40.0) 2 (40.0) > (61.5) 5 (38.5) 11 (84.6) 11 (84.6) Note Values in parentheses are percentages. in five patients, including two lesions in each of four patients and three lesions in one patient found by reader 1. As a result, a total of 53 local recurrence lesions were detected in the 47 patients. The short and long diameters of these 53 lesions ranged from 3 to 26 mm (mean, 7.8 mm) and 5 to 32 mm (mean, 12.0 mm), whereas reader 2 detected 45 local recurrence lesions in 42 patients using protocol C with two lesions in three patients and one lesion in 39 patients. The short and long diameters of these 45 lesions ranged from 3 to 29 mm (mean, 7.9 mm) and 5 to 36 mm (mean, 11.9 mm). The rates of detection of local tumor recurrence by MRI protocol according by lesion size 812 AJR:205, October 2015

7 MRI of Local Recurrence After Prostatectomy TABLE 4: Rate of Detection of Local Tumor Recurrence in the Prostatectomy Bed Stratified by Lesion Size for MRI Protocols Length of MRI Lesion (mm) No. of MRI True-Positive Lesions No. of MRI True-Positive Lesions Protocol A Protocol B Protocol C Protocol D Reader 1 All 80 patients (0) 0 (0) 2 (100) 2 (100) (59.1) 6 (27.3) 22 (100) 22 (100) (75.0) 15 (62.5) 24 (100) 24 (100) (80.0) 5 (100) 5 (100) 5 (100) Total (66.0) 26 (49.1) 53 (100) 53 (100) 49 Patients without second-line therapy (0) 0 (0) 1 (100) 1 (100) (64.3) 4 (28.6) 14 (100) 14 (100) (63.6) 7 (63.6) 11 (100) 11 (100) (66.7) 3 (100) 3 (100) 3 (100) Total (62.1) 14 (48.3) 29 (100) 29 (100) 31 Patients with previous second-line therapy (0) 0 (0) 1 (100) 1 (100) (50.0) 2 (25.0) 8 (100) 8 (100) (84.6) 8 (61.5) 13 (100) 13 (100) (100) 2 (100) 2 (100) 2 (100) Total (70.8) 12 (50) 24 (100) 24 (100) Reader 2 All 80 patients (33.3) 1 (33.3) 3 (100) 3 (100) (85.7) 4 (28.6) 14 (100) 13 (92.9) (75.0) 15 (62.5) 24 (100) 24 (100) (75.0) 4 (100) 4 (100) 4 (100) Total (75.6) 24 (53.3) 45 (100) 44 (97.8) 49 Patients without second-line therapy (0) 0 (0) 1 (100) 1 (100) (90.0) 3 (30.0) 10 (100) 9 (90.0) (71.4) 8 (57.1) 14 (100) 14 (100) (50.0) 2 (100) 2 (100) 2 (100) Total (74.1) 13 (48.1) 27 (100) 26 (96.3) 31 Patients with previous second-line therapy (50.0) 1 (50.0) 2 (100) 2 (100) (75.0) 1 (25.0) 4 (100) 4 (100) (80.0) 7 (70.0) 10 (100) 10 (100) (100) 2 (100) 2 (100) 2 (100) Total (77.8) 11 (61.1) 18 (100) 18 (100) Note Values are numbers of patients with percentages in parentheses. Length of MRI lesion is defined as the maximum lesion size. Detection rate is defined as relative ratio between the number of lesions in the prostatectomy bed for the image protocol and the number of lesions identified with protocol C (T2-weighted and dynamic contrast-enhanced MR images), which was the highest among the imaging protocols for both readers. (long diameter) for each reader are shown in Table 4. For reader 1, for the 53 local recurrence lesions with a long-axis diameter less than 10 mm, the detection sensitivity of protocol B (T2- weighted imaging and DWI) was 25.0% (6/24) and 10 mm or greater, 69.0% (20/29). For reader 2, for 45 local recurrence lesions, the respective values for protocol B were 29.4% (5/17) and 67.9% (19/28). For local recurrence, protocol A (T2-weighted imaging alone) had detection rates of % for lesions with a longaxis diameter less than 10 mm and % for lesions 10 mm and larger. Patients Without Second-Line Therapy The PSA levels of 49 prostatectomy patients who had never undergone salvage therapy before MRI ranged from 0.1 to 10.3 ng/ml (median, 0.41 ng/ml). The per patient sensitivity, specificity, accuracy, and AUC for the four protocols are shown in Table 2. For both readers, the addition of DCE to T2-weighted imaging (protocols C and D) resulted in statistically larger AUCs than for T2-weighted imaging regardless of the use of DWI (protocols A and B). No significant difference was present in the AUCs between protocols A and B or between protocols C and D for either reader. For reader 2, the sensitivity of protocol D was not statically different from that of protocol A (p = ). Otherwise, protocols C and D had significantly better sensitivities and accuracies than protocols A and B for both readers. Protocol A was more sensitive than protocol B for both readers. There was no significant difference in specificity between the protocols for the two readers. A representative case is shown in Figure 2. Patients With Previous Second-Line Therapy The PSA levels of 31 prostatectomy patients who underwent second-line therapy before MRI ranged from 0.1 to 6.3 ng/ml (median, 0.5 ng/ml). Second-line therapy included salvage radiation therapy in 11 patients, androgen deprivation therapy in 10, and both salvage radiation therapy and androgen deprivation therapy in 10 (Table 1). The per-patient sensitivity, specificity, accuracy, and ROC results for the four protocols are shown in Table 2. For both readers, the addition of DCE to T2-weighted imaging (protocols C and D) had statistically higher sensitivity and accuracy and a larger AUC than did T2-weighted imaging regardless of the use of DWI (protocols A and B). For each reader, there was no significant difference in AUCs between protocols A and B or between AJR:205, October

8 Kitajima et al. protocols C and D. There was no significant difference in specificities among the four protocols for either reader. Two representative cases are shown in Figures 3 and 4. Comparison Between Subgroups of Patients With and Without Previous Second-Line Treatment Neither reader had statistically significant differences in sensitivity, specificity, or accuracy between patients without and with previous second-line treatment (p = ). Under protocol C, which had the highest performance, for each reader the ROC curves of the two subgroups were similar regardless of previous second-line treatment (Fig. 5). Discussion Our first major finding was that 3-T endorectal coil T2-weighted and DCE MRI has high sensitivity and specificity in the detection of local tumor recurrence after RP for prostate carcinoma regardless of serum PSA A C level or previous second-line treatment. The second finding was that DWI with ADC mapping adds little value to T2-weighted and DCE MRI with regard to lesion detection. Several investigators [8 12, 17, 18] have reported the usefulness of endorectal coil MRI for detecting local recurrence after RP. The sensitivities have been % and specificities % for a combination of T2-weighted and DCE MRI, in comparison with 48 84% sensitivity and 52 88% specificity for T2-weighted imaging alone. Our results of this multireader study are similar to previously published results and add to the evidence that the addition of DCE to T2-weighted MRI is an important tool for the detection of local recurrence in the prostatectomy bed. MRI is the most accurate imaging modality for clearly defining local recurrence in terms of size, location, and relation to critical structures, such as the sphincters [21]. In our series, serum PSA level did not affect the performance of MRI, B Fig year-old man with increased prostate-specific antigen level of 3.4 ng/ml 10 years after failed radical prostatectomy followed by salvage external beam radiation therapy. Reader 1 assigned a score of 2 for protocols A and B and 4 for C and D. Reader 2 assigned a score of 1 for protocols A and B and 4 for C and D. A, Axial T2-weighted MR image shows 12 5 mm slightly hyperintense softtissue nodule (arrow) in left posterolateral aspect of prostate bed adjacent to left anterior rectal wall. Neither reader detected nodule. B, Apparent diffusion coefficient map shows suspected nodule (arrow) with lower signal intensity, but both readers deemed this finding negative. C, Dynamic contrast-enhanced MR image shows early and avid enhancement of nodule (arrow), suggesting presence of local recurrence. Histopathologic findings obtained at transrectal ultrasound guided biopsy revealed prostate cancer (Gleason 4 + 4). as reported by Panebianco et al. [17]. With the increasing use of more focused, local salvage therapies for recurrent prostate cancer, the detailed information obtained with MRI is becoming invaluable [22]. This is the third reported study of the performance of DWI with ADC mapping in endorectal coil MRI for diagnosing local recurrence after RP. We found that the addition of DWI with ADC mapping to T2-weighted imaging (protocol B) or T2-weighted and DCE MRI (protocol D) did not improve detection compared with T2-weighted imaging alone (protocol A) and T2-weighted and DCE MRI together (protocol C) without DWI with ADC mapping, as reported by Roy et al. [18]. Roy et al. used 3-T endorectal coil MRI and found that the rate of detection of local tumor recurrence was highest for T2-weighted plus DCE MRI (97%), followed by T2-weighted imaging plus DWI plus DCE MRI (94%), T2-weighted imaging plus DWI (65%), and 814 AJR:205, October 2015

9 MRI of Local Recurrence After Prostatectomy Sensitivity Specificity No previous second-line therapy Previous second-line therapy A T2-weighted imaging alone (56%). The detectability of lesions with a combination of T2-weighted imaging and DWI with ADC mapping depended on tumor size. In our series, sensitivity was low (25 29%) for lesions smaller than 10 mm in diameter. The utility of DWI with ADC mapping may be limited by its lower spatial resolution, susceptibility artifacts from surgical clips or gas within the rectum or endorectal coil, and rectal motion. Of interest, protocol A (T2-weighted imaging alone) was more sensitive than protocol B (T2-weighted imaging and DWI with ADC mapping) for both readers. The addition of DWI to T2-weighted imaging led to a falsenegative undercall if no diffusion signal abnormality was present to correspond to an abnormality on the T2-weighted image. Contrary to our findings, Panebianco et al. [17] reported that sensitivity and specificity in depicting local recurrent tumor for combined T2-weighted imaging and DWI were closer to the performance of combined T2-weighted and DCE MRI. It remains to be seen whether the use of DWI with reduced-fov single-shot echo-planar imaging can further improve the detectability of local tumor recurrence after RP. With reference to our clinical practice, we favor higher specificity of MRI in diagnosing local tumor recurrence with confidence scores of 4 and 5 (Table 2). Follow-up with PSA measurement and imaging in 6 12 months is considered appropriate when the MRI findings continue to be indeterminate or negative with a score of 3 or less than 3, because the yield of more invasive transrectal ultrasound guided biopsy is low [23]. This study had several additional points of consideration. First, it was a retrospective study performed at a single institution. Second, histologic examination of all patients would have been ideal but was not performed in all cases for practical and ethical reasons. Third, 197 of 280 patients (70.4%) were excluded because of lack of sufficient reference standard because most patients underwent hormone therapy for nodal or bony metastases. This has potential to introduce selection bias. Fourth, our DCE MRI with fat suppression emphasizes higher spatial and contrast resolution to allow detection of very small enhancing lesions in the prostatectomy bed. Because of this, our sequence had limitations of temporal resolution, and quantitative parametric imaging approaches with pharmacokinetic models could not be applied. Conclusion Performance of MRI at 3 T with an endorectal coil and T2-weighted and DCE technique has good diagnostic performance in the detection of local recurrence of prostate cancer after RP, regardless of serum PSA Specificity No previous second-line therapy Previous second-line therapy Fig. 5 Comparison of ROC curves of protocol C (T2-weighted and dynamic contrast-enhanced MR images) between subgroups of patients without (n = 49) and with (n = 31) previous second-line therapy. AUCs between two subgroups for both readers are similar regardless of previous second-line treatment. A, Graph shows ROC curve for reader 1. AUC is 0.91 both without and with second-line therapy. B, Graph shows ROC curve for reader 2. AUC without second-line therapy is 0.83 and with second-line therapy is Sensitivity level or previous second-line treatment. DWI and ADC mapping have little added value in lesion detection. References 1. Cookson MS, Aus G, Burnett AL, et al. Variation in the definition of biochemical recurrence in patients treated for localized prostate cancer: the American Urological Association prostate guidelines for localized prostate cancer update panel report and recommendations for a standard in the reporting of surgical outcomes. J Urol 2007; 177: Han M, Partin AW, Zahurak M, Piantadosi S, Epstein JI, Walsh PC. Biochemical (prostate specific antigen) recurrence probability following radical prostatectomy for clinically localized prostate cancer. J Urol 2003; 169: Freedland SJ, Mangold LA, Walsh PC, Partin AW. The prostatic specific antigen era is alive and well: prostatic specific antigen and biochemical progression following radical prostatectomy. J Urol 2005; 174: Mottet N, Bellmunt J, Bolla M, et al. EAU guidelines on prostate cancer. Part 2. Treatment of advanced, relapsing, and castration resistant prostate cancer. Eur Urol 2011; 59: Shekarriz B, Upadhyay J, Wood DP Jr, et al. Vesicourethral anastomosis biopsy after radical prostatectomy: predictive value of prostate-specific antigen and pathologic stage. Urology 1999; 54: Leventis AK, Shariat SF, Slawin KM. Local re- B AJR:205, October

10 Kitajima et al. currence after radical prostatectomy: correlation Zhang J, Hricak H. The incremental value of con- my: the role of 3-T diffusion imaging in multi- of US features with prostatic fossa biopsy find- trast-enhanced MRI in the detection of biopsy- parametric magnetic resonance imaging. Eur ings. Radiology 2001; 219: proven local recurrence of prostate cancer after Radiol 2013; 23: Vargas HA, Wassberg C, Akin O, Hricak H. MR radical prostatectomy: effect of reader experience. 18. Roy C, Foudi F, Charton J, et al. Comparative sen- imaging of treated prostate cancer. Radiology AJR 2012; 199: sitivities of functional MRI sequences in detection 2012; 262: Donati OF, Jung SI, Vargas HA, et al. Multipara- of local recurrence of prostate carcinoma after 8. Casciani E, Polettini E, Carmenini E, et al. Endorectal and dynamic contrast-enhanced MRI for detection of local recurrence after radical prostatectomy. AJR 2008; 190: Sciarra A, Panebianco V, Salciccia S, et al. Role of dynamic contrast-enhanced magnetic resonance (MR) imaging and proton MR spectroscopic imaging in the detection of local recurrence after radical prostatectomy for prostate cancer. Eur Urol 2008; 54: Cirillo S, Petracchini M, Scotti L, et al. Endorectal magnetic resonance imaging at 1.5 Tesla to assess local recurrence following radical prostatectomy using T2-weighted and contrast-enhanced imaging. Eur Radiol 2009; 19: Boonsirikamchai P, Kaur H, Kuban DA, Jackson E, Hou P, Choi H. Use of maximum slope images generated from dynamic contrast-enhanced MRI to detect locally recurrent prostate carcinoma after prostatectomy: a practical approach. AJR 2012; 198:[web]W228 W Wassberg C, Akin O, Vargas HA, Shukla-Dave A, metric prostate MR imaging with T2-weighted, diffusion-weighted, and dynamic contrast-enhanced sequences: are all pulse sequences necessary to detect locally recurrent prostate cancer after radiation therapy? Radiology 2013; 268: Morgan VA, Riches SF, Giles S, Deanaley D, desouza NM. Diffusion-weighted MRI for locally recurrent prostate cancer after external beam radiotherapy. AJR 2012; 198: Liauw SL, Pitroda SP, Eggener SE, et al. Evaluation of the prostate bed for local recurrence after radical prostatectomy using endorectal magnetic resonance imaging. Int J Radiat Oncol Biol Phys 2013; 85: Panebianco V, Sciarra A, Lisi D, et al. Prostate cancer: 1HMRS-DCEMR at 3T versus [(18)F] choline PET/CT in the detection of local prostate cancer recurrence in men with biochemical progression after radical retropubic prostatectomy (RRP). Eur J Radiol 2012; 81: Panebianco V, Barchetti F, Sciarra A, et al. Prostate cancer recurrence after radical prostatecto- radical prostatectomy or external-beam radiotherapy. AJR 2013; 200:[web]W361 W Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44: Sella T, Schwartz LH, Swindle PW, et al. Suspected local recurrence after radical prostatectomy: endorectal coil MR imaging. Radiology 2004; 231: Woodrum DA, Kawashima A, Karnes RJ, et al. Magnetic resonance imaging-guided cryoablation of recurrent prostate cancer after radical prostatectomy: initial single institution experience. Urology 2013; 82: Linder BJ, Kawashima A, Woodrum DA, et al. Early localization of recurrent prostate cancer after prostatectomy by endorectal coil magnetic resonance imaging. Can J Urol 2014; 21: AJR:205, October 2015