| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Genitourinary Imaging |
1 From the Departments of Radiology (L.W., H.N.C., S.C.E., H.H.) and Urology (M.M., M.W.K., P.T.S.), Memorial Sloan-Kettering Cancer Center, 1275 York Ave, New York, NY 10021. From the 2002 RSNA scientific assembly. Received July 10, 2003; revision requested September 16; revision received December 19; accepted January 13, 2004. Supported by National Institutes of Health grant R01 CA76423. Address correspondence to L.W. (e-mail: wang6@mskcc.org).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: In this cohort study, 344 consecutive patients with biopsy-proved prostate cancer underwent endorectal MR imaging prior to surgery; 216 of these patients also underwent MR spectroscopic imaging. MR images were interpreted by 10 attending radiologists. The likelihood of ECE was scored retrospectively on the basis of MR imaging reports. Clinical variables included serum prostate-specific antigen (PSA) level, Gleason score, clinical stage of tumor, greatest percentage of cancer in all core biopsy specimens, percentage of cancer-positive core specimens in all core biopsy specimens, and presence of perineural invasion. For data analysis, receiver operating characteristic (ROC) curves and univariate and multivariate logistic regression analyses were used. Jackknife analysis was used for prediction of probability from a model that included clinical variables as tested comparatively with a model that included the clinical variables plus endorectal MR imaging findings. A difference with P < .05 was considered significant.
RESULTS: At univariate analysis, all variables were associated with ECE. At ROC univariate analysis, endorectal MR imaging findings had the largest area under the ROC curve. At multivariate analysis, serum PSA level, percentage of cancer in all core biopsy specimens, and endorectal MR imaging findings (P = .001, P = .001, and P < .001, respectively) were predictors of ECE. Areas under ROC curve for two models, with and without endorectal MR imaging findings, were 0.838 and 0.772, respectively (P = .022).
CONCLUSION: A model containing endorectal MR imaging findings has a significantly larger area under the ROC curve than a model containing only clinical variables; thus, endorectal MR imaging findings add incremental value in the prediction of ECE.
© RSNA, 2004
Index terms: Magnetic resonance (MR), coils, 844.121411, 844.121419 • Magnetic resonance (MR), spectroscopy, 844.12145 • Neoplasms, staging, 844.32 • Prostate, biopsy, 844.1261 • Prostate neoplasms, 844.32
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Various methods for the prediction of ECE have been proposed. The established presurgical clinical variables for the prediction of pathologic stage are clinical tumor stage, which includes digital rectal examination findings; serum PSA level; and Gleason score (4–6). Nomograms that are based on presurgical data were introduced to help evaluate patient risk for ECE, seminal vesicle invasion, and metastasis to lymph nodes, as well as 5-year freedom from cancer recurrence. These models have high accuracy for prediction with either aggressive or clinically indolent tumors. The assessment of additional presurgical variables to enhance the specificity and sensitivity of the present models is of great interest.
The aim of this study was to assess the incremental value of endorectal magnetic resonance (MR) imaging findings in addition to clinical variables for prediction of ECE in patients with prostate cancer.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Mean patient age was 57.5 years (range, 32–74 years). None of the patients received neoadjuvant hormonal or radiation therapy prior to surgery. Tissue diagnosis of prostate cancer was established with biopsy specimens in all patients. Clinical and MR imaging data were recorded retrospectively from the patients’ medical records by one author (L.W.). Data in the same patient cohort are reported in the study by Mullerad et al (7).
Imaging and Image Interpretation
Endorectal MR imaging and hydrogen 1 MR spectroscopic imaging were performed with a 1.5-T whole-body MR imaging unit (Signa Horizon; GE Medical Systems, Milwaukee, Wis). The examination was performed with patients in the supine position. A body coil was used for excitation, and a pelvic phased-array coil (GE Medical Systems) combined with a commercially available balloon-covered expandable endorectal coil (Medrad, Pittsburgh, Pa) was used for signal reception. Transverse T1-weighted and spin-echo MR images were obtained from the aortic bifurcation to the symphysis pubis with the following parameters: repetition time msec/echo time msec, 700/8; section thickness, 5 mm; intersection gap, 1 mm; field of view, 24 cm; matrix, 256 x 192; and frequency direction, transverse (to prevent obstruction of the pelvic node from endorectal coil motion artifact). One signal was acquired.
Transverse and coronal thin-section high-spatial-resolution T2-weighted fast spin-echo MR images of the prostate and seminal vesicle were obtained with the following parameters: 5,000/96 (effective); echo train length, 16; section thickness, 3 mm; intersection gap, 0 mm; field of view, 14 cm; matrix, 256 x 192; frequency direction, anteroposterior (to prevent obstruction of the prostate from endorectal coil motion artifact); and number of signals acquired, three.
MR spectroscopic imaging was performed by using point-resolved spectroscopic voxel excitation (8), band-selective inversion with gradient dephasing water and lipid suppression (9), and spatial encoding with chemical shift imaging (10) at 6.25-mm resolution in all three dimensions (left-right, anterior-posterior, superior-inferior dimensions). Parameters were 1,000/130, and imaging time was 17 minutes.
Data processing was performed at a workstation (Sun Ultra 10; Sun Microsystems, Mountain View, Calif) and included 2-Hz lorentzian spectral apodization; four-dimensional Fourier transform; and automated frequency, phase, and baseline correction (11). Spectral data were zero filled to 3.1-mm resolution in the superior-inferior dimension and overlaid on corresponding transverse T2-weighted MR images. Peak areas were calculated by using numeric integration. To provide a noise measurement, we calculated the SD of the MR signal intensity in a region of the spectrum containing only noise. Metabolite peak areas were then normalized with respect to the noise SD to yield an approximate signal-to-noise ratio.
MR images were interpreted by 10 body MR imaging radiologists (including H.H., S.C.E.) during their clinical assignment to the MR imaging service. The readers’ experience in interpretation of clinical MR images since fellowship ranged from 4 to more than 15 years. Official pretreatment MR image readings were used for the data analysis. There was no initial meeting or training to establish the criteria for ECE. Rather, radiologists made their determinations on the basis of their own continuing medical training and knowledge of previously described MR imaging features of ECE. The diagnostic criteria used by the radiologists to determine ECE on endorectal MR images included irregular capsular bulge, periprostatic fat infiltration, obliteration of the rectoprostatic angle, and asymmetry or direct involvement of the neurovascular bundles (12). MR spectroscopic imaging results, when available, were provided for all readers. The extent to which MR spectroscopic imaging data were used in image interpretation differed among radiologists. On the basis of the radiologists’ written reports, one author (L.W.) retrospectively scored the likelihood for ECE with a five-point scale as follows: score 1, no ECE; score 2, probably no ECE (cannot be ruled out though there is no clear evidence of it); score 3, possible ECE (a lesion is suspected of demonstrating ECE); score 4, probable ECE (a lesion is highly suspected of demonstrating ECE); score 5, definite ECE.
Histologic Analysis
Histologic analysis reports at core biopsy were evaluated for Gleason score, greatest percentage of cancer in all core biopsy specimens, percentage of cancer-positive core specimens in all core biopsy specimens (the number of cancer-positive core specimens divided by the total number of core biopsy specimens), and presence of perineural invasion (PNI). The greatest percentage of cancer was determined in each patient by examining each core biopsy specimen and dividing the length of the core specimen tissue with cancer by the whole core specimen length; the core specimen with the highest percentage of cancer defined the patient’s greatest percentage of cancer. This parameter has been previously described by Rubin et al (13) and Bismar et al (14).
Specimens removed at radical prostatectomy were examined in the pathology department at our institution, as previously described by Yossepowitch et al (15). In short, specimens were fixed in formalin, with the external surface of the right and left sides inked with two colors. The apical prostate was truncated perpendicular to the prostatic urethra and was subsequently sectioned as slices parallel to the prostatic urethra. The bladder neck margin was obtained by sampling portions of soft tissue at the junction of the rough prostatic capsule and smooth bladder neck or the most proximal portion of the submitted specimen that corresponded to the anatomic bladder neck. The remaining prostate was completely transected at 3–5-mm intervals in a plane perpendicular to the urethra. The final pathologic report following surgery was used to determine the presence of ECE. The presence of cancer cells beyond the capsular margin was used as the definition of ECE.
Statistical Analysis
Both univariate and multivariate analyses were performed for all clinical and imaging variables tested. Predictor variables that we tested included serum PSA level, Gleason score, clinical stage of tumor, greatest percentage of cancer in all core biopsy specimens, percentage of cancer-positive core specimens in all core biopsy specimens, and presence of PNI. With receiver operating characteristic (ROC) univariate analysis, the incremental value of MR spectroscopic imaging findings was assessed by comparing findings in the group of 216 patients who underwent both endorectal MR imaging and MR spectroscopic imaging with those in the group of 128 patients who underwent only endorectal MR imaging.
We evaluated the area under the ROC curve for each variable. In addition, we compared two models: one that contained all clinical variables plus endorectal MR imaging findings and another that contained only the clinical variables. To judge the value of endorectal MR imaging findings as a marker, the predictions analyzed with jackknife methods from these models were compared for their ability to predict ECE. By using bootstrapping for bias correction, a P value was derived to test for a difference in the predictive ability of these two models (16). A difference with P < .05 was considered significant. Software programs (SAS, version 8.2, SAS Institute, Cary, NC; S-Plus, version 2000, Insightful, Seattle, Wash; and Stata, version 7.0, Stata, College Station, Tex) were used for data analysis.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Findings in another investigation (22) and in our study indicate that serum PSA levels are associated with ECE with both univariate and multivariate analysis (P 
.001). In our study, Gleason score and clinical stage of tumor were significantly associated with ECE only with univariate analysis (P < .001 and P = .005, respectively). With multivariate analysis, Gleason score had borderline significance as a predictor in both models, with and without endorectal MR imaging findings (P = .074 and .057, respectively), while clinical stage of tumor was not significant. Similar findings were reported by Horiguchi et al (23). They demonstrated that MR imaging findings, along with Gleason score (categorized as 6 or 
7) and PSA density, were significantly associated with ECE with multivariate logistic regression analysis (P < .01).
Several features demonstrated at prostate biopsy were extensively examined as predictors of final pathologic stage (eg, greatest percentage of cancer in all core biopsy specimens, percentage of cancer-positive core specimens in all core biopsy specimens, and presence of PNI). Both Ravery et al (24) and Freedland et al (25) found percentage of cancer at core biopsy to be predictive of pathologic stage and risk of biochemical recurrence when they used multivariate analysis. Linson et al (26), however, found that only the percentage of cancer-positive core specimens was predictive of biochemical recurrence with multivariate analysis and therefore concluded that pursuing percentage of cancer in cancer-positive core specimens is an unnecessary and time-consuming task. In our study, with univariate analysis, both greatest percentage of cancer in all core biopsy specimens and percentage of cancer-positive core specimens in all core biopsy specimens were significant for the prediction of ECE (P < .001 for both values). When multivariate analysis was used, however, only greatest percentage of cancer in all core biopsy specimens remained a significant predictor (P < .001). This finding was not affected when results of endorectal MR imaging were integrated into the analysis.
Although there is controversy in regard to the value of PNI for the prediction of ECE, some urologists still use the finding of PNI in surgical planning (27). It has been suggested that the ipsilateral nerve bundle should be excised during radical prostatectomy when PNI is present (27,28). Investigators in several studies (20,29,30) address the utility of the finding of PNI in the prediction of ECE. Our results agree with the findings of Vargas et al (20) and of Egan and Bostwick (31) and show that the presence of PNI is a significant predictor of ECE with univariate analysis. In addition, progression of disease has also been examined.
In a regression analysis, Stone et al (32) found that the presence of PNI in a biopsy specimen correlated with lymph node metastasis at radical prostatectomy (P = .04). D’Amico et al (33) and de la Taille et al (34) demonstrated that presence of PNI significantly correlates with PSA recurrence after radical prostatectomy. O’Malley et al (35), however, were not able to show a significant difference in PSA recurrence when they compared patients with PNI and patients without it who were treated with radical prostatectomy (29,36). With our multivariate analysis, presence of PNI was not significant for the prediction of ECE (P = .987).
The quest for a better diagnostic test that can help differentiate between advanced and localized disease and assist a physician in treatment planning led to the evaluation of endorectal MR imaging in prostate cancer localization and staging. The routine use of MR imaging for presurgical evaluation of prostate cancer is controversial. The high incidence of the disease combined with the high cost of the test might burden the health care system with additional expenses unless its use prevents unnecessary surgery or aids in treatment planning and results in better outcomes (37). D’Amico et al (38) concluded that although MR imaging findings add significant predictive value (ie, prediction of the risk of developing biochemical failure following radical prostatectomy in 20% of patients), this result did not justify the routine use of the technique. Furthermore, endorectal MR imaging results demonstrated high interobserver variability, which limited the widespread use of the technique (39,40).
It has been suggested that treatment decisions should not be altered because of either endorectal MR imaging or transrectal ultrasonographic findings (41). In our study of the comparison of the value of endorectal MR imaging findings and clinical and histologic variables, endorectal MR imaging findings had a high negative predictive value and a high positive predictive value (83.8% and 74.5%, respectively). When assessed with univariate analysis, endorectal MR imaging findings demonstrated excellent prediction of ECE (P < .001). In our study, the addition of MR spectroscopic imaging did not result in a significant improvement in image interpretation (P < .206). Previously published data support the value of MR spectroscopic imaging in the detection and staging of prostate cancer (12). The incremental value of MR spectroscopic imaging was greater for the less experienced reader (12). Notably, MR spectroscopic technology in our institution was introduced in 2000, and there was a steep learning curve in the acquisition and interpretation of MR spectroscopic imaging data during the past 4 years.
With multivariate analysis, endorectal MR imaging findings, PSA level, and greatest percentage of cancer in all core biopsy specimens were all predictors of ECE (P < .001). Because we recognized that results of such an analysis are insufficient for judgment of endorectal MR imaging findings as a new marker (16), we examined the incremental effect of endorectal MR imaging findings on predictive accuracy. When we compared the two models, one with and one without endorectal MR imaging findings, at multivariate analysis, we found that the area under the ROC curve of 0.838 for the model with endorectal MR imaging findings was significantly greater than that of 0.772 for the model without endorectal MR imaging findings (P = .022). These data demonstrate the incremental value of endorectal MR imaging findings. In addition, endorectal MR imaging findings are spatially localized, and therefore, unlike clinical variables, they have the potential to allow tailored treatment modifications (12,42–44).
The strengths and limitations of our study are drawn from the fact that it was conducted in the routine clinical setting of the radiology department. Ten body MR imaging radiologists interpreted the images as part of their routine clinical assignments. There were differences in training and experience in prostate imaging among the readers, and their experience in interpretation of clinical MR images ranged from 4 to more than 15 years since fellowship. This may have resulted in a lack of data uniformity and in interobserver variability in the diagnosis of ECE. Since each image was interpreted by one reader, we could not assess the effect of interobserver variability on the accuracy of prediction of ECE at endorectal MR imaging. In a recently published study of tumor staging with meta-analysis at MR imaging, Engelbrecht et al (45) demonstrated considerable heterogeneity between endorectal MR imaging studies; they concluded that further studies are needed to establish the effect of the reader’s experience, as well as the effect of the clinical information given to the reader, as they hypothesized that these factors may cause considerable differences in staging accuracy between the studies. A contrast material–enhanced study was not used in our protocol, though recent data suggest that such a study might improve accuracy in tumor staging (46).
With consideration of all these factors, our results show that endorectal MR imaging findings have remarkable strength in the prediction of ECE. Although further multicenter confirmatory study findings would be helpful, we suggest that endorectal MR imaging findings play an important role in the evaluation of prostate cancer and the prediction of ECE.
In conclusion, endorectal MR imaging findings are significant presurgical predictors of ECE in patients with prostate cancer, and they add incremental value to clinical variables.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: ECE = extracapsular extension, PNI = perineural invasion, PSA = prostate-specific antigen, ROC = receiver operating characteristic
Author contributions: Guarantors of integrity of entire study, H.H., M.W.K.; study concepts, H.H.; study design, H.H., M.W.K.; literature research, L.W.; clinical studies, H.H., P.T.S., S.C.E.; data acquisition, L.W.; data analysis/interpretation, H.H., M.W.K., S.C.E.; statistical analysis, M.W.K., H.N.C.; manuscript preparation, H.H., M.W.K., M.M.; manuscript definition of intellectual content and revision/review, H.H.; manuscript editing, H.H., M.M., L.W., S.C.E., P.T.S.; manuscript final version approval, H.H., P.T.S., M.W.K.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. C. Westphalen, F. V. Coakley, A. Qayyum, M. Swanson, J. P. Simko, Y. Lu, S. Zhao, P. R. Carroll, B. M. Yeh, and J. Kurhanewicz Peripheral Zone Prostate Cancer: Accuracy of Different Interpretative Approaches with MR and MR Spectroscopic Imaging Radiology, December 1, 2007; 246(1): 177 - 184. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Wang, Y. Mazaheri, J. Zhang, N. M. Ishill, K. Kuroiwa, and H. Hricak Assessment of Biologic Aggressiveness of Prostate Cancer: Correlation of MR Signal Intensity with Gleason Grade after Radical Prostatectomy Radiology, December 1, 2007; 246(1): 168 - 176. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Shukla-Dave, H. Hricak, C. Moskowitz, N. Ishill, O. Akin, K. Kuroiwa, J. Spector, M. Kumar, V. E. Reuter, J. A. Koutcher, et al. Detection of Prostate Cancer with MR Spectroscopic Imaging: An Expanded Paradigm Incorporating Polyamines Radiology, November 1, 2007; 245(2): 499 - 506. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. N. Bloch, E. Furman-Haran, T. H. Helbich, R. E. Lenkinski, H. Degani, C. Kratzik, M. Susani, A. Haitel, S. Jaromi, L. Ngo, et al. Prostate Cancer: Accurate Determination of Extracapsular Extension with High-Spatial-Resolution Dynamic Contrast-enhanced and T2-weighted MR Imaging Initial Results Radiology, October 1, 2007; 245(1): 176 - 185. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. Hricak, P. L. Choyke, S. C. Eberhardt, S. A. Leibel, and P. T. Scardino Imaging Prostate Cancer: A Multidisciplinary Perspective Radiology, April 1, 2007; 243(1): 28 - 53. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Wang, J. Zhang, L. H. Schwartz, H. Eisenberg, N. M. Ishill, C. S. Moskowitz, P. Scardino, and H. Hricak Incremental Value of Multiplanar Cross-Referencing for Prostate Cancer Staging with Endorectal MRI Am. J. Roentgenol., January 1, 2007; 188(1): 99 - 104. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. J. Choi, J. K. Kim, N. Kim, K. W. Kim, E. K. Choi, and K.-S. Cho Functional MR Imaging of Prostate Cancer RadioGraphics, January 1, 2007; 27(1): 63 - 75. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Wang, H. Hricak, M. W. Kattan, H. N. Chen, K. Kuroiwa, H. F. Eisenberg, and P. T. Scardino Prediction of Seminal Vesicle Invasion in Prostate Cancer: Incremental Value of Adding Endorectal MR Imaging to the Kattan Nomogram Radiology, December 1, 2006; 242(1): 182 - 188. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. Akin, E. Sala, C. S. Moskowitz, K. Kuroiwa, N. M. Ishill, D. Pucar, P. T. Scardino, and H. Hricak Transition Zone Prostate Cancers: Features, Detection, Localization, and Staging at Endorectal MR Imaging Radiology, June 1, 2006; 239(3): 784 - 792. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Wang, H. Hricak, M. W. Kattan, L. H. Schwartz, S. C. Eberhardt, H.-N. Chen, and P. T. Scardino Combined endorectal and phased-array MRI in the prediction of pelvic lymph node metastasis in prostate cancer. Am. J. Roentgenol., March 1, 2006; 186(3): 743 - 748. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Sala, S. C. Eberhardt, O. Akin, C. S. Moskowitz, C. N. Onyebuchi, K. Kuroiwa, N. Ishill, M. J. Zelefsky, J. A. Eastham, and H. Hricak Endorectal MR Imaging before Salvage Prostatectomy: Tumor Localization and Staging Radiology, January 1, 2006; 238(1): 176 - 183. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Wang, H. Hricak, M. W. Kattan, H.-N. Chen, P. T. Scardino, and K. Kuroiwa Prediction of Organ-confined Prostate Cancer: Incremental Value of MR Imaging and MR Spectroscopic Imaging to Staging Nomograms Radiology, December 12, 2005; (2005) 2382041905. [Abstract] [Full Text]
|
|
|
|
|
|
J. J. Futterer, S. W. T. P. J. Heijmink, T. W. J. Scheenen, G. J. Jager, C. A. Hulsbergen-Van de Kaa, J. A. Witjes, and J. O. Barentsz Prostate Cancer: Local Staging at 3-T Endorectal MR Imaging--Early Experience Radiology, December 1, 2005; 238(1): 184 - 191. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H Hricak MR imaging and MR spectroscopic imaging in the pre-treatment evaluation of prostate cancer Br. J. Radiol., October 1, 2005; 78(Special_Issue_2): S103 - S111. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Mullerad, H. Hricak, L. Wang, H.-N. Chen, M. W. Kattan, and P. T. Scardino Prostate Cancer: Detection of Extracapsular Extension by Genitourinary and General Body Radiologists at MR Imaging Radiology, July 1, 2004; 232(1): 140 - 146. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
