Using Gray-Scale and Color and Power Doppler Sonography to Detect Prostatic Cancer Ethan J. Halpern 1 Stephen E. Strup 2 OBJECTIVE. We performed a prospective study to assess gray-scale and color and power Doppler sonography for the detection of prostatic cancer and to determine the impact of operator experience. SUBJECTS AND METHODS. Four radiologists with prior experience using gray-scale and Doppler imaging and four urologists with prior experience limited to gray-scale imaging performed sextant biopsies on 251 patients. Each biopsy site was prospectively scored for gray-scale and Doppler abnormality. RESULTS. Cancer was detected in 211 biopsy sites from 85 patients. Overall agreement between sonographic findings and biopsy results as measured with the kappa statistic was minimally superior to chance (κ = 0.12 for gray-scale, κ = 0.11 for color Doppler, κ 0.09 for power Doppler). With respect to gray-scale diagnosis of cancer, the performance of radiologists (κ = 0.12) and urologists (κ = 0.13) was similar. With respect to power Doppler, the performance of radiologists (κ = 0.09) was superior to that of urologists (κ = 0.03, p < 0.002). Among patients with at least one positive biopsy for cancer, foci of increased power Doppler flow detected by a radiologist were 4.7 times more likely to contain cancer than adjacent tissues without flow. CONCLUSION. Gray-scale and Doppler imaging did not reveal prostatic cancer with sufficient accuracy to avoid sextant biopsy. Power Doppler may be useful for targeted biopsies when the number of biopsy passes must be limited. There is benefit from increased operator experience with Doppler imaging, but there is no demonstrable benefit of power Doppler over conventional color Doppler sonography. Received June 24, 1999; accepted after revision August 10, 1999. 1 Department of Radiology, Jefferson Prostate Center, Thomas Jefferson University, 132 S. 10th St., Philadelphia, PA 19107-5244. Address correspondence to E. J. Halpern. 2 Department of Urology, Jefferson Prostate Center, Thomas Jefferson University, Philadelphia, PA 19107-5244. AJR 2000;174:623 627 0361 803X/00/1743 623 American Roentgen Ray Society T he number of new cases of prostatic cancer in the United States in 1999 is estimated at 179,300 with 37,000 resulting in death [1]. Assuming one third of prostate biopsies reveal cancer, more than 500,000 prostate biopsies are performed annually. Although most clinicians use sonography to guide the biopsy, differences of opinion exist with regard to the role of sonography [2]. Only 20% of urologists perform targeted biopsy on the basis of sonographic findings [3]. Cancer of the prostate classically presents as a hypoechoic lesion [4] but can appear echogenic or isoechoic [5]. Although increased cancer detection has been reported with color Doppler sonography [6 8], the combined sensitivity of gray-scale and color Doppler imaging is insufficient to preclude sextant biopsy [9 11]. Power Doppler is more sensitive to slow flow and is less angledependent than color Doppler imaging [12]. Results of several small studies have suggested that power Doppler sonography may be useful in detection of prostatic cancer [13,14]. To our knowledge, no large series has evaluated power Doppler imaging for the detection of prostatic cancer, and no study has directly compared color and power Doppler sonography of the prostate. Our study correlates findings on grayscale and color and power Doppler imaging with results from sextant biopsy. The focus is to evaluate the diagnostic accuracy of power Doppler sonography and the impact of operator experience on the Doppler examination. Subjects and Methods Our study population consisted of 251 patients who underwent sonographically guided biopsy of the prostate between October 1997 and June 1999. Patients ranged in age from 37 to 87 years, with a AJR:174, March 2000 623
Halpern and Strup mean age of 64.6 years. The serum level of prostate-specific antigen (PSA) was available for 223 patients and was elevated (>4 ng/dl) in 190. A PSA value of greater than 10 was found in 53 patients. The average PSA value was 8.7 ng/dl. All patients were examined with the 6.5EC10 endfire probe using the Sonoline Elegra system (Siemens; Issaquah, WA). For gray-scale imaging, the center probe frequency was 6.0 MHz with a dynamic range of 55 db. For color and power imaging, the center probe frequency was 4.0 MHz with a dynamic range of 30 db, pulse repetition frequency was 868 Hz, and wall filter was set to low. Color and power gain were adjusted as follows: gain was increased until clutter was observed and then reduced just enough to remove clutter from the prostate. Transrectal examination consisted of a standard sequence of axial images from base to apex, followed by sagittal images from right to left. Sextant biopsy of the outer gland was performed on all patients. Six independent biopsy specimens from the outer gland were obtained from each patient. Individual biopsy specimens were obtained from the base, mid gland, and apex on each side of the gland. When an abnormality was visualized on gray-scale or Doppler imaging, the corresponding sextant biopsy was directed to the site of the abnormality. An 18- gauge core biopsy system (ASAP; Medi-tech, Boston Scientific, Watertown, MA) was used. The imaging protocol changed in 1998 when the departments of radiology and urology formed a joint prostate imaging center. In the 3 months before formation of this joint center, 41 patients were examined by one of four experienced radiologists using grayscale and color Doppler sonography. These 41 patients were our baseline for color Doppler imaging. During 1998, four experienced urologists were added to the prostate center staff. Although each of the radiologists and urologists was experienced with gray-scale sonography, the urologists had no prior experience with Doppler imaging. Between January 1998 and June 1999, 210 patients were examined with gray-scale and power Doppler sonography. Sonographic findings were recorded at the base, mid portion, and apex of the gland on both the right and left sides. Gray-scale was classified as normal, indeterminate, or abnormal at each biopsy site. A normal site was homogeneous in texture with no focal contour bulge. A site was classified as abnormal if a definite mass, either echogenic or hypoechoic, was present. Doppler flow was subjectively classified as absent, minimal, or increased at each biopsy site. Imaging findings were scored prospectively at the time of the examination. For statistical analysis, indeterminate gray-scale findings were classified as abnormal and minimal Doppler flow was classified as increased. A kappa value was computed to document the agreement between sonographic findings and pathology. The kappa value quantifies the level of agreement relative to that which might be expected by chance. A kappa value of 1 corresponds to perfect disagreement, whereas a kappa value of +1 corresponds to perfect agreement. A kappa value of 0 corresponds to chance agreement. Although the kappa value is generally used to assess agreement between two observers, we have expanded its application to assess agreement between sonographic findings and needle biopsy results. The advantage of the kappa value over the traditional measures of diagnostic accuracy lies in its ability to assess observer performance relative to random agreement. To determine the value of sonography in selecting the biopsy site in patients with cancer, conditional logistic regression was applied. Conditional logistic regression allows matching of biopsy sites within each patient and accounts for the lack of independence of multiple sites within an individual patient. Statistical analyses were performed for the entire patient population and were repeated for various subsets of patients. All statistical computations were performed with Stata 6.0 software (Stata, College Station, TX). Results Malignant tissue was found in a total of 211 biopsy sites of 85 patients. Gleason scores ranged from 3 to 10, with a mean value of 6.7. The most common Gleason scores were 6, 7, and 8, which accounted for 38%, 33%, and 14% of all malignancies, respectively. Grayscale and color or power Doppler abnormalities were often associated with the presence of prostatic cancer (Figs. 1 and 2). With respect to the detection of malignant lesions (Table 1), gray-scale sonography yielded 44.1% sensitivity with 73.6% specificity (κ = 0.12), whereas Doppler imaging (color and power) yielded 27.0% sensitivity with 77.1% specificity (κ = 0.03). Thirty-five of the 211 lesions were detected with both gray-scale and Doppler imaging, but 96 lesions were missed by both these techniques. Even among patients with a PSA level of greater than 10 ng/dl, grayscale detected only 40% of the malignant lesions, whereas Doppler imaging detected 10%. Conditional logistic regression was performed to evaluate the use of gray-scale and Doppler imaging (color and power) in patients with proven prostatic cancer. Each variable showed a significant positive correlation with cancer when it was included as the only independent variable (gray-scale: odds ratio = 1.9, p = 0.014; Doppler: odds ratio = 3.7, p < 0.001). When both variables were simultaneously included, a positive correlation was seen for gray-scale (odds ratio = 1.4, 95% confidence interval [CI] = 0.8 2.5) and Doppler flow (odds ratio = 3.2, 95% CI = 1.5 6.9), but only Doppler flow was significant (p = 0.003). The distribution of Gleason scores among malignant specimens with sufficient tissue for grading is summarized in Table 2. The mean Gleason score was 6.5 among foci with no associated gray-scale abnormality and 7.0 among foci with gray-scale findings. The mean Gleason score was 6.6 among cancerous foci with no Doppler flow and 7.0 among cancerous foci with Doppler flow. The chi-square test for trend was significant (χ = 6.3, p = 0.012) for tumor grade as a function of gray-scale abnormality but was not significant for tumor grade versus Doppler flow (χ = 2.9, p = 0.086). Among 41 patients in the baseline group who were examined with color Doppler imaging, 48 foci with cancer were found in 16 patients. Among 210 subsequent patients examined with power Doppler imaging, 163 foci with cancer were found in 69 patients. Average Gleason Fig. 1. Adenocarcinoma of prostate involving left base and left mid gland in 73-year-old man. Gleason score was 7 (3 + 4). Transverse gray-scale sonogram shows large hypoechoic mass on left (arrows) with associated bulge of posterolateral prostatic contour. Power Doppler shows increased flow within and around hypoechoic tumor. 624 AJR:174, March 2000
Sonographic Detection of Prostatic Cancer Fig. 2. Adenocarcinoma of the prostatic base with bilateral involvement in 61-year-old man. Gleason score was 6 (3 + 3). A, Transverse color sonogram shows increased parenchymal flow on right. Left-sided flow is predominantly in capsular vessels and neurovascular bundle. B, Transverse power sonogram also shows increased flow on right side. A B score in both groups was 6.7. The populations were similar in terms of age and PSA level. Color Doppler imaging yielded 14.6% sensitivity and 93.9% specificity (κ = 0.11), whereas power Doppler imaging yielded 30.7% sensitivity and 74.0% specificity (κ = 0.03). Conditional logistic regression with both gray-scale and Doppler sonography as independent variables showed a positive correlation of cancer with color flow (odds ratio = 3.9, p = 0.15) and power Doppler (odds ratio = 2.9, p = 0.015). The lack of significance for color Doppler was related to small sample size in the baseline group. A comparison of results for radiologists and urologists is provided in Table 3. With respect to gray-scale imaging, the sensitivity of radiologists was 44.4% with a specificity of 70.5% (κ = 0.12). The sensitivity of urologists was 37.5% with a specificity of 79.8% (κ = 0.13). With power Doppler imaging, the sensitivity of radiologists was 27.3% with a specificity of 83.9% (κ = 0.09). The sensitivity of urologists was 35.9% with a specificity of 59.3% (κ = 0.03). Agreement between power Doppler and pathology results was significantly better for the radiologists than for the urologists (p < 0.002). With the exception of power Doppler interpretation by the urologists, the remaining kappa values in this study were significant (p < 0.05). Conditional logistic regression analysis AJR:174, March 2000 625
Halpern and Strup Gray-Scale and Doppler TABLE 1 Detection of Prostatic Cancer by Biopsy Site in 251 Patients Sonographic Findings Benign Malignant Gray-scale sonography Normal findings 953 118 Abnormal findings 342 93 Doppler sonography No flow 998 154 Increased flow 297 57 TABLE 2 Sonographic Findings Gray-Scale and Doppler Detection of Prostatic Cancer by Biopsy Site, Stratified by Gleason Score Gleason Score 3 4 5 6 7 8 9 10 Gray-scale sonography Normal findings 2 0 5 51 27 10 6 0 Abnormal findings 2 1 3 18 32 16 7 1 Doppler sonography No flow 4 0 6 53 42 19 7 0 Increased flow 0 1 2 16 17 7 6 1 Gray-scale and Doppler sonography Normal findings 2 0 4 42 22 8 5 0 Gray-Scale and Power Doppler Detection of TABLE 3 Prostatic Cancer Stratified by Radiologists Versus Urologists Sonographic Findings Benign Malignant Radiologists Gray-scale sonography Normal findings 463 55 Abnormal findings 194 44 Power Doppler sonography No flow 551 72 Increased flow 106 27 Urologists Gray-scale sonography Normal findings 351 40 Abnormal findings 89 24 Power Doppler sonography No flow 261 41 Increased flow 179 23 of gray-scale and power Doppler was repeated for the subgroup of patients examined by radiologists, showing a positive correlation for both gray-scale (odds ratio = 1.8, p = 0.14) and power Doppler (odds ratio = 4.7, p = 0.007). Discussion On the most basic level, sonography may be used to direct biopsy samples from all parts of the gland. A more advanced role of sonography is to select and target biopsy sites with cancerous foci. In this role, sonography might be used to limit the number of biopsy sites per patient or even to limit the number of patients subjected to biopsy. Although our results do not provide a direct comparison of targeted and sextant biopsy strategies, our results do indicate that gray-scale and Doppler evaluation fail to identify many malignant lesions found on sextant biopsy. Because the trend at many centers is to obtain a larger number of biopsy samples (i.e., saturation biopsies), a strategy that relies only on a limited number of targeted biopsies will miss some malignant lesions. Is the effort required for diagnostic sonography of the prostate justified? Even among patients with an elevated PSA level ( 10 ng/ml) who might be expected to have larger tumor burdens, sonography showed low sensitivity. Furthermore, sonography failed to detect 35 of 99 cancerous foci with a Gleason score of 7 or higher (Table 2). The results of a recent study comparing sextant and targeted biopsy in 194 patients revealed that additional malignant lesions were not detected with targeted biopsy [15]. Nonetheless, regression analysis suggests that gray-scale and Doppler sonography are useful to select biopsy sites in patients with cancer. Thus, although normal sonographic findings should not preclude biopsy, regions with gray-scale or Doppler abnormality may be preferentially sampled in patients with limited tolerance or at high risk for multiple biopsies. The urologists in our study had no prior experience with color or power Doppler sonography. Although the urologists performed as well as the radiologists with gray-scale imaging, kappa analysis suggests that the radiologists were more accurate with power Doppler imaging (p < 0.002). Clearly, there is a learning curve for power Doppler imaging. However, even for the radiologists, there was no diagnostic advantage to using power Doppler (κ = 0.09) over color Doppler (κ = 0.11) in the diagnosis of prostatic cancer. An association has been shown between increased microvessel density in prostatic cancer and the presence of metastases [16], the stage of disease [17 19], and disease-specific survival [20, 21]. Indeed, increased color flow correlates with tumor stage and grade as well as with risk of recurrence after treatment [22]. If microvascular density is increased in prostatic cancer, why do color and power Doppler imaging fail to detect malignant lesions? The key to this discrepancy may be related to the size and distribution of microvessels in prostatic cancer. Although there are more vessels in malignant prostate tissue, the distribution of these microvessels is more uniform [23] and these vessels are smaller [24]. The total intravascular volume in malignant tissue may not be much greater than that found in benign tissue [21]. Because microvessels are below the limit of resolution of color and power Doppler imaging, larger feeder vessels, which supply vascular beds of similar volumes in benign and malignant prostate tissue, are predominantly visualized with color and power Doppler imaging. Several limitations of this study should be acknowledged. Imaging findings were prospectively interpreted at the time of the study, but the examining physician was not blinded to clinical and laboratory information. The number of patients in the baseline color Doppler imaging group was small. All examinations were performed with the Sonoline Elegra system. Although this system provides high-quality imaging, our results may not necessarily be generalized to other instruments or imaging techniques. In conclusion, although there are structural differences in the architecture of benign and malignant prostate tissue, gray-scale and color and power Doppler sonography do not adequately distinguish these features. Future research on diagnostic imaging of prostatic cancer should address parameters that might distinguish known microscopic differences in glandular and microvessel patterns. References 1. Landis SH, Murray T, Bolden S, Wingo PA. Cancer statistics, 1999. CA Cancer J Clin 1999;49:8 31 2. Bree RL. The role of color Doppler and staging biopsies in prostate cancer detection. Urology 1997;49[suppl 3A]:31 34 3. Plawker MW, Fleisher JM, Vapnek EM, Macchia RJ. Current trends in prostate cancer diagnosis and staging among United States urologists. J Urol 1997;158:1853 1858 4. Rifkin MD, Dahnert W, Kurtz AB. State of the art: endorectal sonography of the prostate gland. AJR 1990;154:691 700 5. Dahnert WF, Hamper UM, Eggleston JC, et al. Prostatic evaluation by transverse sonography with histo- 626 AJR:174, March 2000
Sonographic Detection of Prostatic Cancer pathologic correlation: echogenic appearance of early carcinoma. Radiology 1986;158:97 102 6. Rifkin MD, Sudakoff GS, Alexander AA. Prostate: techniques, results and potential applications of color Doppler US scanning. Radiology 1993;186: 509 513 7. Sudakoff GS, Smith R, Vogelzang NJ. Color Doppler imaging and transrectal sonography of the prostate fossa after radical prostatectomy: early experience. AJR 1996;167:883 888 8. Lavoipierre AM, Snow RM, Frydenberg M. Prostatic cancer: role of color Doppler imaging in transrectal sonography. AJR 1998;171:205 210 9. Kelly IMG, Lees WR, Rickards D. Prostate cancer and the role of color Doppler US. Radiology 1993;189:153 156 10. Newman JS, Bree RL, Rubin JM. Prostate cancer: diagnosis with color Doppler sonography with histologic correlation of each biopsy site. Radiology 1995;195:86 90 11. Cornud F, Belin X, Piron D, et al. Color Dopplerguided prostate biopsies in 591 patients with an elevated serum PSA level: impact on Gleason score for non-palpable lesions. Urology 1997;49: 709 715 12. Bude RO, Rubin JM. Power Doppler sonography. Radiology 1996;200:21 23 13. Cho JY, Kim SH, Lee SE. Diffuse prostatic lesions: role of color Doppler and power Doppler ultrasonography. J Ultrasound Med 1998;17:283 287 14. Okihara K, Kojima M, Naya Y, Iida A, Watanabe M, Watanabe H. Ultrasonic power Doppler imaging for prostatic cancer: a preliminary report. Tohoku J Exp Med 1997;182:277 281 15. Blake MA, Barish MA, Fenlon HF, Pomeroy O, Kuligowska E. Transrectal ultrasound (TRUS) detection of prostate cancer: comparison of sextant versus targeted biopsy. J Ultrasound Med 1999;18:S21 16. Weidner N, Carroll PR, Flax J, et al. Tumor angiogenesis correlates with metastasis in invasive prostate carcinoma. Am J Pathol 1993;143:401 409 17. Fregene TA, Khanuja PS, Noto AC, et al. Tumorassociated angiogenesis in prostate cancer. Anticancer Res 1993;13:2377 2382 18. Brawer MK, Deering RE, Brown M, et al. Predictors of pathologic stage in prostate carcinoma. The role of neovascularity. Cancer 1994;73:678 687 19. Bostwick DG, Wheeler TM, Blute M, et al. Optimized microvessel density analysis improves prediction of cancer stage from prostate needle biopsies. Urology 1996;48:47 57 20. Lissbrant IF, Stattin P, Damber JE, et al. Vascular density is a predictor of cancer-specific survival in prostatic carcinoma. Prostate 1997;33:38 45 21. Borre M, Offersen BV, Nerstrom B et al. Microvessel density predicts survival in prostate cancer patients subjected to watchful waiting. Br J Cancer 1998;78:940 944 22. Ismail M, Petersen RO, Alexander AA, et al. Color Doppler imaging in predicting the biologic behavior of prostate cancer: correlation with disease-free survival. Urology 1997;50:906 912 23. Kay PA, Robb RA, Bostwick DG. Prostate cancer microvessels: a novel method for three-dimensional reconstruction and analysis. Prostate 1998; 37:270 277 24. Louvar E, Littrup PJ, Goldstein A, et al. Correlation of color flow in the prostate with tissue microvascularity. Cancer 1998;83:135 140 AJR:174, March 2000 627
Halpern and Strup 628 AJR:174, March 2000