Radiology
Genitourinary Imaging Rajpal Dhingsa, MD Aliya Qayyum, MD Fergus V. Coakley, MD Ying Lu, PhD Kirk D. Jones, MD Mark G. Swanson, PhD Peter R. Carroll, MD Hedvig Hricak, MD, PhD John Kurhanewicz, PhD
Index terms: Diagnostic radiology, observer performance Magnetic resonance (MR), spectroscopy Prostate neoplasms, 844.32 Prostate neoplasms, MR, 844.121411, 844.12145 Published online 10.1148/radiol.2301021562 Radiology 2004; 230:215–220 Abbreviations: AFROC ⫽ alternative free-response ROC PSA ⫽ prostate-specific antigen ROC ⫽ receiver operating characteristic 1
From the Departments of Radiology (R.D., A.Q., F.V.C., J.K., M.G.S., Y.L.), Pathology (K.D.J.), and Urology (P.R.C.), University of California San Francisco, 505 Parnassus Avenue, San Francisco, CA 94143-0628; and Department of Radiology, Memorial Sloan-Kettering Cancer Center, New York (H.H.). Received November 25, 2002; revision requested January 28, 2003; final revision received May 28; accepted June 12. Supported by NIH grant IRGICA76423– 0IRI. Address correspondence to A.Q. (e-mail:
[email protected])
Author contributions: Guarantor of integrity of entire study, R.D.; study concepts, F.C., A.Q., H.H.; study design, F.C., A.Q., R.D., J.K.; literature research, R.D., A.Q.; clinical studies, F.C., A.Q.; data acquisition, R.D., A.Q., K.J.; data analysis/interpretation, F.C., A.Q., Y.L., K.J., J.K., M.G.S.; statistical analysis, Y.L.; manuscript preparation, definition of intellectual content, editing, revision/review, and final version approval, all authors ©
RSNA, 2004
Prostate Cancer Localization with Endorectal MR Imaging and MR Spectroscopic Imaging: Effect of Clinical Data on Reader Accuracy1 PURPOSE: To determine the effect of digital rectal examination findings, sextant biopsy results, and prostate-specific antigen (PSA) levels on reader accuracy in the localization of prostate cancer with endorectal magnetic resonance (MR) imaging and MR spectroscopic imaging. MATERIALS AND METHODS: This was a retrospective study of 37 patients (mean age, 57 years) with biopsy-proved prostate cancer. Transverse T1-weighted, transverse high-spatial-resolution, and coronal T2-weighted MR images and MR spectroscopic images were obtained. Two independent readers, unaware of clinical data, recorded the size and location of suspicious peripheral zone tumor nodules on a standardized diagram of the prostate. Readers also recorded their degree of diagnostic confidence for each nodule on a five-point scale. Both readers repeated this interpretation with knowledge of rectal examination findings, sextant biopsy results, and PSA level. Step-section histopathologic findings were the reference standard. Logistic regression analysis with generalized estimating equations was used to correlate tumor detection with clinical data, and alternative free-response receiver operating characteristic (AFROC) curve analysis was used to examine the overall effect of clinical data on all positive results. RESULTS: Fifty-one peripheral zone tumor nodules were identified at histopathologic evaluation. Logistic regression analysis showed awareness of clinical data significantly improved tumor detection rate (P ⬍ .02) from 15 to 19 nodules for reader 1 and from 13 to 19 nodules for reader 2 (27%–37% overall) by using both size and location criteria. AFROC analysis showed no significant change in overall reader performance because there was an associated increase in the number of false-positive findings with awareness of clinical data, from 11 to 21 for reader 1 and from 16 to 25 for reader 2. CONCLUSION: Awareness of clinical data significantly improves reader detection of prostate cancer nodules with endorectal MR imaging and MR spectroscopic imaging, but there is no overall change in reader accuracy, because of an associated increase in false-positive findings. A stricter definition of a true-positive result is associated with reduced sensitivity for prostate cancer nodule detection. ©
RSNA, 2004
Accurate localization of prostate cancer within the gland is of increasing clinical importance due to the development of disease-targeted (or potentially disease targeted) ablative therapies such as interstitial brachytherapy, intensity-modulated radiation therapy, highintensity focused ultrasonography, and cryosurgery (1). In addition, tumor localization has been related to the risk of tumor recurrence after prostatectomy, with a higher risk when surgical margins are positive at the base than when margins are positive at the apex (2). Combined endorectal magnetic resonance (MR) imaging and MR spectroscopic imag215
Radiology
ing has a reported sensitivity and specificity for prostate cancer sextant localization of 73% and 80%, respectively (3–5). Combined endorectal MR imaging and MR spectroscopic imaging has an accuracy rate similar to that of biopsy for prostate cancer sextant localization, and it is more accurate than biopsy of the prostate apex (5). In these previous studies of MR imaging and MR spectroscopic imaging, image interpretation was performed by readers unaware of clinical data such as digital rectal examination findings, sextant biopsy results, and prostate-specific antigen (PSA) levels. In practice, urologists routinely integrate this information into their evaluation of prostate cancer extent and aggressiveness, as exemplified by the so-called Partin Tables (6). In studies performed in areas other than prostate imaging, clinical data have generally been shown to improve radiologic evaluation (7–13), but, to our knowledge, the influence of clinical data on interpretation of MR images and MR spectroscopic images has not been investigated. Therefore, we undertook this study to determine the effect of digital rectal examination findings, sextant biopsy results, and PSA level on reader accuracy in the localization of prostate cancer with endorectal MR imaging and MR spectroscopic imaging.
MATERIALS AND METHODS Patients This was a retrospective single-institution study (University of California San Francisco). We included all patients who underwent radical prostatectomy over a 1-year period and who underwent preoperative endorectal MR imaging and MR spectroscopic imaging of the prostate (n ⫽ 37). Contraindication to MR imaging was the only exclusion criteria. Patients were referred to undergo endorectal MR imaging and MR spectroscopic imaging after the diagnosis of prostate cancer had been established with biopsy results, and these patients were recruited as part of an ongoing National Institutes of Health study to investigate the use of MR imaging in prostate cancer. Results of a study in the same patient group were reported in a recent article on evaluating MR imaging and MR spectroscopic imaging measurement of prostate cancer tumor volume (14). Our study received institutional Committee on Human Research approval, and written informed consent was obtained from all patients. 216
䡠
Radiology
䡠
January 2004
The mean patient age was 57 years (range, 43–75 years). The median tumor Gleason score was 6 (range, 5–9). The mean PSA level at diagnosis was 6.8 ng/mL (range, 2.2–21.9 ng/mL). The mean interval from biopsy to endorectal MR imaging and MR spectroscopic imaging was 8 weeks (range, 0 –59 weeks). The mean interval from endorectal MR imaging and MR spectroscopic imaging to surgery was 6 weeks (range, 0 –23 weeks). None of the patients received preoperative hormonal or radiation therapy.
MR Imaging Technique MR imaging was performed by using a 1.5-T whole-body MR imager (Signa Echospeed 1.5; GE Medical Systems, Milwaukee, Wis). Patients were imaged in a supine position by using the body coil for excitation and a pelvic phased-array coil (GE Medical Systems) in combination with a commercially available ballooncovered expandable endorectal coil (Medrad, Pittsburgh, Pa) for signal reception. Transverse spin-echo T1-weighted MR images were obtained from the aortic birfucation to the symphysis pubis by using the following parameters: 7,000/8 (repetition time msec/echo time msec), section thickness of 5 mm, intersection gap of 1 mm, field of view of 24 cm, matrix of 256 ⫻ 192, transverse frequency direction (to prevent obscuration of pelvic nodes by endorectal-coil motion artifact), and one signal acquired. Transverse thin-section high-spatial-resolution and coronal T2-weighted fast spin-echo MR images of the prostate and seminal vesicles were obtained by using the following parameters: 5,000/96 (effective), echo train length of 16, section thickness of 3 mm, intersection gap of 0 mm, field of view of 14 cm, matrix of 256 ⫻ 192, anteroposterior frequency direction (to prevent obscuration of the prostate by endorectal-coil motion artifact), and three signals acquired. All MR images were routinely postprocessed to compensate for the reception profile of the endorectal and pelvic phased-array coils. After the transverse T2-weighted MR images were reviewed, a spectroscopic MR imaging volume was selected by an experienced spectroscopist (J.K. or M.G.S.) in conjunction with a technologist to maximize coverage of the prostate while minimizing the inclusion of periprostatic fat and rectal air. Three-dimensional MR spectroscopic imaging data were acquired by using a water- and lipid-suppressed double spin-echo point-resolved spatially localized spectroscopy sequence, also known
as PRESS, that was optimized for the quantitative detection of both choline and citrate. Water and lipid suppression was achieved by using band selective inversion with gradient dephasing, or BASING (15). Outer voxel saturation pulses were also employed to eliminate susceptibility artifacts from periprostatic fat and rectal air. Data sets were acquired as 16 ⫻ 8 ⫻ 8 phase-encoded spectral arrays (1,024 voxels) with a nominal spectral resolution of 0.24 – 0.34 cm3, 1,000/130, and a 17-minute acquisition time. The total examination time was 1 hour, which included coil placement and patient positioning.
Spectroscopic Data Evaluation Spectroscopic MR imaging data were overlaid on the corresponding transverse T2-weighted MR images and evaluated in consensus by two experienced spectroscopists (J.K. and M.G.S.; 14 and 6 years experience, respectively) to determine those voxels that were suitable for analysis. Voxels were considered suitable if they consisted of at least 75% peripheral zone tissue, did not include tissue surrounding the urethra or ejaculatory ducts, had a signal-to-noise ratio greater than 5:1, and were not spectroscopically contaminated by insufficient water or fat suppression. For each useable voxel, the ratio of choline plus creatine to citrate was calculated. Useable peripheral zone voxels with ratios of choline plus creatine to citrate that were greater than 3 SDs above the normal mean value were considered to indicate malignancy and were designated “C” for cancer. Useable voxels with a ratio of choline plus creatine to citrate that was greater than 2 SDs from the normal mean value were considered to indicate possible malignancy and were designated “P” for possible cancer (3). These designations were marked on a grid overlaid on the corresponding transverse T2-weighted MR images.
Image Interpretation Two radiologists (F.V.C. and A.Q.; 7 and 4 years experience, respectively, in prostate MR imaging and MR spectroscopic imaging interpretation) independently reviewed all MR images in conjunction with the MR spectroscopic image overlays. Readers performed two interpretations. In the initial interpretation, readers were aware of the diagnosis of prostate cancer but were unaware of all other clinical data. Readers identified peripheral zone tumors as nodules of low Dhingsa et al
Radiology
signal intensity on T2-weighted images, as clusters of voxels designated as malignant or possibly malignant on spectroscopic MR images, or as both. In practice, endorectal MR imaging and MR spectroscopic imaging are often synergistic with respect to tumor identification; for example, an equivocal endorectal MR imaging finding may be considered more or less suspicious because of the presence or absence (respectively) of correlative MR spectroscopic imaging findings (3,4). By using their best estimate of tumor location, which was based on the combination of endorectal MR imaging and MR spectroscopic imaging, readers marked tumor location on a standardized diagram of the prostate. Since prostate cancer is often multifocal, more than one tumor location could be marked for each patient. For each possible tumor nodule, readers recorded the maximum transverse diameter (in millimeters) and their level of confidence. Level of confidence was rated on a five-point scale according to the following criteria: very uncertain (score of 1), uncertain (score of 2), intermediate (score of 3), confident (score of 4), or very confident (score of 5). In the second interpretation, readers were aware of the diagnosis of prostate cancer and were also aware of digital rectal examination findings, sextant biopsy results (Gleason score, location of cancer-containing cores, and length of tumor in cancer-containing cores), and PSA level. To minimize learning bias, the order of images was altered for the second interpretation and a period of at least 2 weeks was allowed to elapse between successive interpretations. In the second interpretation, readers again noted the location, size, and level of confidence for each potential peripheral zone tumor nodule.
Histopathologic Analysis Specimens removed at radical prostatectomy were coated with India ink and fixed in 10% buffered formaldehyde. Transverse step sections were obtained at 3– 4-mm intervals in a plane perpendicular to the long axis of the prostate. An experienced pathologist (K.D.J.) recorded the location of all peripheral zone tumor nodules on a standardized diagram of the prostate and recorded the maximum transverse diameter of all tumor nodules. A preliminary analysis of the first 10 specimens showed no difference in the pre- and postfixation weights of the specimens removed at prostatectomy and, therefore, no correction factor was used to correct for tumor shrinkage during fixVolume 230
䡠
Number 1
ation. The pathologist also recorded the presence or absence of extracapsular extension and seminal vesicle invasion.
Statistical Analysis The unit of analysis was the peripheral zone tumor nodule. In the analysis, each nodule recorded by the readers on the schematic prostate diagram was compared by a radiologist (R.D.) with the true location of peripheral zone tumor on the pathologic tumor map prepared by the pathologist. Tumors detected with MR imaging were considered true-positive findings at histopathologic correlation when tumor was present on the pathologic map within the same region that was considered tumorous at imaging, and the sizes of the tumors were comparable. Processing of prostatectomy specimens is associated with tissue shrinkage reported to range between 18% and 33% (16,17). To provide reasonable allowance for differences in registration, as well as morphology between imaging and histopathologic evaluation, the tumors were considered to be of comparable sizes if the maximum transverse diameter measured at MR imaging was within the range of 50%–150% of the maximum transverse diameter at histopathologic evaluation. All other nodules detected at MR imaging were considered false-positive findings. The statistical analysis consisted of two parts. First, we examined tumor detection with MR imaging and MR spectroscopic imaging by comparing the proportion of true-positive results obtained with and without clinical data among all tumor nodules confirmed at histopathologic evaluation. The mean sizes of the true-positive and false-negative lesions were recorded. The mean sizes of additional true-positive and false-positive lesions with availability of clinical data were recorded, with mean confidence level for both readers. Logistic regression analysis with generalized estimating equations was used to account for the clustering effects of multiple tumor nodules in the same patient, and two independent readers evaluated the same nodules. P ⬍ .05 was considered to show statistical significance. Second, we used alternative free-response receiver operating characteristic (AFROC) curve analysis to examine the overall effect of clinical data on all positive calls, including changes in proportion of true-positive results, the chance of at least one false-positive result in a patient, and the level of confidence (18).
AFROC is preferable to receiver operating characteristic (ROC) in the analysis of the presence and location of disease within a given patient. In this setting, a localizing diagnostic test generates only true-positive (“hits”), false-positive (“overcalls”), and false-negative (“misses”) responses. Readers interpreting localizing diagnostic tests for a disease or tumor do not generally circle areas considered “disease free.” This lack of true-negative results limits the use of ROC analysis in this setting, which is primarily useful in the analysis of whether disease is present or absent in a group of patients. AFROC analysis can be used to evaluate disease localization in patients known to have disease. AFROC analysis plots the fraction of nodules with truepositive ratings v against the probability of a false-positive image P(FPI) as a function of the reader’s confidence. A higher confidence level is associated with lower sensitivity for v but lower false-positive image reporting. As confidence level is reduced, both sensitivity and false-positive image reporting are increased. The area under the curve in AFROC analysis combines true-positive, false-positive, and false-negative ratings and level of confidence, and it represents a measure of the trade-off between detecting truepositive versus false-positive lesions (percentage of true-positive findings before the first false-positive finding) over a range of thresholds (18). In an ROC analysis, the probability of correct classification of observer response is equivalent to the area under the ROC curve and should be greater than 0.5 to ensure correct classification does not occur by chance alone. There is no clear equivalent value for the area under the curve in an AFROC analysis. Chakraborty and Winter (18) reported values for the area under the AFROC curve ranging from 35.1% to 45.2%. We did not perform statistical tests to compare the area under the AFROC curves because there is no consensus on the appropriate method for handling of the correlated data in an AFROC analysis. Statistical calculations were performed by using statistical software package SAS 8.2 (SAS Institute, Cary, NC), S-plus (MathSoft, Seattle, Wash), and ROCKIT (University of Chicago, Ill).
RESULTS A total of 51 peripheral zone tumor nodules were identified at histopathologic evaluation. Within the peripheral zone,
Prostate Cancer Localization with MR Imaging
䡠
217
Radiology
21 patients had a single nodule, 12 patients had two nodules each, and two patients had three nodules each. Two patients had tumors confined to the transition zone. Transition zone tumor nodules were not included in the study because, as standard practice at our institution, MR spectroscopic imaging is performed only in the peripheral zone because of limitations of volume coverage and variability of transition zone spectra in benign prostatic hypertrophy. Thirtythree patients had organ-confined tumors, and four patients had extracapsular tumor extension. None of the patients had seminal vesicle invasion. The mean histopathologic tumor nodule diameter was 14 mm (range, 4 –28 mm). Reader 1 correctly identified 15 tumor nodules without knowing clinical data and 19 nodules with the clinical data. The corresponding numbers for reader 2 were 13 and 19, respectively (Fig 1). According to the two readers, the mean histopathologic size of true-positive lesions was 16 mm (range, 5–20 mm), and the mean histopathologic size of the falsenegative lesions was 10 mm (range, 4 –28 mm). The mean histopathologic size of the additional lesions identified correctly by the readers with availability of clinical data was 11 mm (range, 5–21 mm) (mean level of confidence for the two readers, 3.5). Use of logistic regression analysis confirmed that awareness of digital rectal examination findings, sextant biopsy results, and PSA level significantly improved the tumor detection rate from 27% to 37% (P ⬍ .02). No other variable was found to be significantly associated with detection rate by using multivariate logistic regression analysis, although a near significant association was found between histopathologic tumor size and tumor detection rate (P ⫽ .06) (Table 1). Reader 1 identified 11 false-positive nodules without clinical data and 21 with clinical data. The corresponding numbers for reader 2 were 16 and 25, respectively (Fig 2). The mean size of the false-positive lesions on MR images with availability of clinical data was 8 mm for reader 1 and 7 mm for reader 2 (range, 3–19 mm), with a mean level of confidence of 3.2 and 3.8, respectively. AFROC analysis demonstrated a tendency for the area under the AFROC curve to decrease with reader awareness of digital rectal examination findings, sextant biopsy results, and PSA level from 0.69 to 0.50 for reader 1 and from 0.49 to 0.43 for reader 2 (Table 2), although statistical significance could not be appropriately evaluated (Fig 3). 218
䡠
Radiology
䡠
January 2004
TABLE 1 Univariate Logistic Regression Analysis for Sensitivity of Tumor Detection Variable
Odds Ratio
95% CI
P Value
Reader (1 vs 2) Method (without vs with clinical data) Age (per year increase) Pathologic size (per cm increase) PSA level (per 1 unit) Gleason score (ⱕ6 vs ⬎6)
1.09 0.64 0.96 1.10 1.00 0.93
0.86, 1.40 0.45, 0.91 0.90, 1.03 1.00, 1.20 0.88, 1.14 0.34, 2.57
.47 .01 .29 .06 .98 .89
Figure 1. Transverse T2-weighted MR image (5,000/96 [effective]) obtained in a 49-year-old man with a PSA level of 4.5 ng/mL and positive finding after transrectal biopsy, which indicated a Gleason grade 6 tumor at the right apex. Image shows a tumor focus (arrow) at the right apex, which was only identified by both readers after clinical data were made available.
DISCUSSION In our study, reader awareness of digital rectal examination findings, sextant biopsy results, and PSA level significantly increased the rate of detection of peripheral zone tumor nodules. However, this increase in true-positive findings was offset by an increase in false-positive findings, so that there was no overall improvement in reader accuracy, which was demonstrated by AFROC analysis. Berbaum et al (9) reported superior detection of abnormalities on chest radiographs with correct categorical prompts for specific abnormalities and few false-positive results with plausible but incorrect prompts. Leslie et al (10) studied the effect of clinical data on reader accuracy in interpretation of computed tomographic
(CT) images. In 19 of 50 cases, availability of clinical data resulted in alteration of the CT report; accuracy improved in 10 and declined in five of the 19 cases. In three of the five cases with reduced reader accuracy, the clinical data were found to be incorrect. The improvement results reported in most articles in the general radiology literature (7–13) on evaluating the effect of clinical data on imaging interpretation are discordant with our results. This may be explained by an inherent limitation in the clinical assessment of prostate cancer related to sampling error at biopsy and localization of biopsy cores (5). Radiologists are susceptible to suggestion. Knowledge of clinical data is likely to result in increased reporting of abnormalities, such as reporting of bronchial wall thickening on Dhingsa et al
Radiology Figure 2. Transverse T2-weighted MR image (5,000/96 [effective]) obtained in a 66-year-old man with a PSA level of 4.8 ng/mL and positive finding after transrectal biopsy, which indicated a Gleason grade 6 tumor at the left mid-gland. Image shows a false-positive tumor focus (arrow) at the left mid-gland, which was only identified by both readers after clinical data were made available.
TABLE 2 Area Under the AFROC Curve for Tumor Detection Unaware of Clinical Data
Reader 1 2
Aware of Clinical Data
0.69 (0.49, 0.85) 0.50 (0.34, 0.66) 0.49 (0.31, 0.68) 0.43 (0.15, 0.83)
Note.—Data in parentheses are 95% CIs.
pediatric chest radiographs during a bronchiolitis epidemic (19,20). Knowledge of clinical data can also result in underestimation of abnormalities, as was highlighted by results of one study, which demonstrated a shift in reported bone age toward the normal range in 15% of cases with knowledge of patient age compared with reported bone age at blind interpretation (21). Inaccurate clinical data have been shown to be detrimental to radiology reporting (10) and may also be a pitfall to overall performance in interpretation of the prostate by using endorectal MR imaging and MR spectroscopic imaging. In our study we used AFROC analysis, which is a statistical technique that we believe may be more appropriate than conventional ROC analysis. In an experVolume 230
䡠
Number 1
Figure 3. Graph shows AFROC curves of interpretation of MR images and MR spectroscopic images with and without clinical data. AFROC analysis plots the fraction of nodules with true-positive ratings (v) against the probability of a false-positive image (P[FPI]). The area under the curve in AFROC analysis is a measure of overall test accuracy that combines true-positive and false-positive results and level of confidence, and it can be considered to represent the probability that the first tumor detected in a patient is a true-positive rather than a false-positive result. W ⫽ with; W/O ⫽ without.
iment with ROC analysis, the radiologist classifies images as normal or abnormal with a confidence rating (typically a fivepoint scale). The ROC plot of the observer data is a plot of the true-positive fraction against the false-positive fraction, which is obtained by cumulating the responses. The area under the ROC curve is an objective measure of imaging system performance. However, diagnostic tasks are often not equivalent to simple yes or no decisions, and they usually require additional specifications such as the location of the abnormality. Tumor localization is important in evaluating prostate cancer, but in ROC analysis we cannot take location information into account, which makes the analysis susceptible to ambiguous scoring (18). Per sextant ROC analysis of the prostate has been used, such that each sextant is classified as normal or abnormal, to enable indirect interpretation of tumor nodule location. However, a general feature in studies involving MR imaging evaluation of prostate cancer is that patients without cancer are excluded, so that there are no true-negative results, and interpretation of true-negative sextants is difficult due to the lack of anatomic boundaries
between sextants. Per sextant ROC analysis of the prostate is also limited by the lack of size concordance as part of the definition of a true-positive result, which is accounted for in the AFROC analysis (18). In our study, the detection rate of 37% for prostate cancer tumor nodules by using endorectal MR imaging and MR spectroscopic imaging with knowledge of digital rectal examination, sextant biopsy results, and PSA level was lower than the reported sensitivities of 69%– 80% in prior studies (3,5,14,22). The data from our patient group were recently published in an article on a study in which tumor localization alone was used to evaluate prostate cancer detection, which demonstrated a sensitivity ranging between 69% and 76% by using MR imaging and MR spectroscopic imaging (14). Our stricter definition of a true-positive result, in which we require concordance of both location and size (diameter at imaging within 50%–150% of the true diameter) between imaging and histopathologic examinations, is likely to have reduced our sensitivity in comparison with the results of other studies, in which per sextant analysis was used (3,5).
Prostate Cancer Localization with MR Imaging
䡠
219
Radiology
The mean tumor nodule diameter in our study was 14 mm (range, 4 –28 mm), and this small nodule size may have also been a contributing factor to the low detection rate. The definition of a true-positive result using a 50%–150% ratio of nodule size measured at MR imaging to size measured at histopathologic evaluation was selected to allow for shrinkage of the prostatectomy specimen, as well as MR imaging and histopathologic misregistration. Jager et al (22) reported 25% overestimation of tumor volume on endorectal MR images in 43% (19 of 44) of prostate cancer nodules and 25% underestimation of tumor volume in 18% (eight of 44) of prostate cancer nodules. This 25% discrepancy in tumor volume correlates with a 63% discrepancy in diameter. The relationship between prostate cancer detection on MR images and tumor size is likely complex. The magnitude of metabolic changes detected by using MR spectroscopic imaging is reportedly related to the Gleason score (23), with only a small increase in choline levels detected in low-grade tumors. The median Gleason score in our study was 6 (range, 5–9), which may have also contributed to low detection rates. Although small foci of low-grade tumors may be more difficult to visualize, these are not generally considered clinically important in the context of prostate cancer. A further explanation for the low sensitivity may be the trend toward imaging in a patient population with earlier stages of disease. The rate of extracapsular extension reported in prior studies was 38%, (4) which suggests a higher volume of disease and a higher stage, compared with only 11% extracapsular extension in our patient group. In a recent study, in which factors predictive of pT3 tumor and positive endorectal MR images were evaluated, a 69% sensitivity and 95% specificity of MR imaging was reported in patients with at least three positive sextant biopsies, a palpable tumor, and/or a PSA level greater than 10 ng/mL (24). The accuracy of MR imaging and MR spectroscopic imaging appears to be related to the extent of disease, which should be taken into consideration when requesting these examinations. There are several limitations to our study. This was a retrospective study of patients known to have prostate cancer, and the patient group has been previously studied, which may confer a bias toward tumor detection. The number of patients in our study was small (n ⫽ 37), which may have been a contributing fac-
220
䡠
Radiology
䡠
January 2004
tor to the marginal significance of the association between histopathologic tumor size and detection rate. We did not use a consensus interpretation; instead, the histopathologic tumor maps were prepared by a pathologist but were correlated with schematic diagrams of the prostate by a radiologist, and this may have resulted in under- or overestimation of true-positive results. However, this study was directed at the effect of clinical data on reader interpretation, and the results of image interpretation both with and without clinical data were compared with the histopathologic tumor maps by using the same technique. In conclusion, awareness of clinical data significantly improves reader detection of prostate cancer nodules by using endorectal MR imaging and MR spectroscopic imaging, but there is no overall change in reader accuracy because there is an associated increase in false-positive findings. A stricter definition of a truepositive result is associated with reduced sensitivity for prostate cancer nodule detection. References 1. Carroll PR, Presti JC Jr, Small E, Roach M III. Focal therapy for prostate cancer 1996: maximizing outcome. Urology 1997; 49(suppl 3A):84 –94. 2. Blute M, Bostwick DG, Bergstralh EJ, et al. Anatomic site-specific positive margins in organ-confined prostate cancer and its impact on outcome after radical prostatectomy. Urology 1997; 50:733–739. 3. Scheidler J, Hricak H, Vigneron DB, et al. Prostate cancer: localization with threedimensional proton MR spectroscopic imaging– clinicopathologic study. Radiology 1999; 213:473– 480. 4. Yu KK, Scheidler J, Hricak H, et al. Prostate cancer: prediction of extracapsular extension with endorectal MR imaging and three-dimensional proton MR spectroscopic imaging. Radiology 1999; 213: 481– 488. 5. Wefer AE, Hricak H, Vigneron DB, et al. Sextant localization of prostate cancer: comparison of sextant biopsy, magnetic resonance imaging and magnetic resonance spectroscopic imaging with step section histology. J Urol 2000; 164:400 – 404. 6. Partin AW, Mangold LA, Lamm DM, Walsh PC, Epstein JI, Pearson JD. Contemporary update of prostate cancer staging nomograms (Partin Tables) for the new millennium. Urology 2001; 58:843– 848. 7. Schreiber MN. The clinical history as a factor in roentgenogram interpretation. JAMA 1963; 185:399 – 401. 8. Hessel SJ, Hermann PG. The value of searching films without specific preconceptions. Invest Radiol 1985; 20:100 – 107. 9. Berbaum KS, Franken EA Jr, Dorfman DD, et al. Tentative diagnoses facilitate the de-
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20. 21.
22.
23.
24.
tection of diverse lesions in chest radiographs. Invest Radiol 1986; 21:532–539. Leslie A, Jones AJ, Goddard PR. The influence of clinical information on the reporting of CT by radiologists. Br J Radiol 2000; 73:1052–1055. Potchen EJ, Gard JW, Lazar P, Lahaie P, Andary M. The effect of clinical history data on chest film interpretation: direction or distraction (abstr). Invest Radiol 1979; 14:404. Eldevik OP, Dugstad G, Orrison WW, Haughton VM. The effect of clinical bias on the interpretation of myelography and spinal computed tomography. Radiology 1982; 145:85– 89. Good BC, Cooperstein LA, DeMarino GB, et al. Does knowledge of the clinical history affect the accuracy of chest radiograph interpretation? AJR Am J Roentgenol 1990; 154:709 –712. Coakley FV, Kurhanewicz J, Lu Y, et al. Prostate cancer tumor volume: measurement with endorectal MR and MR spectroscopic imaging. Radiology 2002; 223: 91–97. Males RG, Vigneron DB, Star-Lack J, et al. Clinical application of BASING and spectral/spatial water and lipid suppression pulses for prostate cancer staging and localization by in vivo 3D 1H magnetic resonance spectroscopic imaging. Magn Reson Med 2000; 43:17–22. Partin AW, Epstein JI, Cho KR, Gittlesohn AM, Walsh PC. Morphometric measurement of tumor volume and percent of gland involvement as predictors of pathological stage in clinical stage B prostate cancer. J Urol 1989; 141:341–345. McNeal JE, Bostwick DG, Kindrachuck RA, Redwine EA, Freiha FS, Stamey TA. Patterns of progression in prostate cancer. Lancet 1986; 1:60 – 63. Chakraborty DP, Winter LH. Free-response methodology: alternate analysis and a new observer-performance experiment. Radiology 1990; 174:873– 881. Babcook CJ, Norman GR, Coblentz CL. Effect of clinical history on the interpretation of chest radiographs in childhood bronchiolitis. Invest Radiol 1993; 28:214 – 217. Griscom NT. A suggestion: look at the images first, before you read the history. Radiology 2002; 223:9 –10. Berst MJ, Dolan L, Bogdanowicz MM, Stevens MA, Chow S, Brandser EA. Effect of knowledge of chronologic age on the variability of pediatric bone age determined using the Greulich and Pyle standards. AJR Am J Roentgenol 2001; 176: 507–510. Jager GJ, Ruijter ET, van de Kaa CA, et al. Local staging of prostate cancer with endorectal MR imaging: correlation with histopathology. AJR Am J Roentgenol 1996; 166:845– 852. Kurhanewicz J, Vigneron DB, Males RG, Swanson M, Yu KK, Hricak H. The prostate: MR imaging and spectroscopy— present and future. Radiol Clin North Am 2000; 38:115–138. Cornud F, Flam T, Chauveinc L, et al. Extraprostatic spread of clinically localized prostate cancer: factors predictive of pT3 tumor and of positive endorectal MR imaging examination results. Radiology 2002; 224:203–210.
Dhingsa et al
