- Email: [email protected]
Evaluation of Transrectal Ultrasound in the Early Detection of Prostate Cancer
0022-534 7/89/1424-1008$02.00/0 Vol. 142, October
THE JOURNAL OF UROLOGY Copyright© 1989 by AMERICAN UROLOGICAL ASSOCIATION, INC.
Printed in U.S.A.
EVALUATION OF TRANSRECTAL ULTRASOUND IN THE EARLY DETECTION OF PROSTATE CANCER H. BALLENTINE CARTER, ULRIKE M. HAMPER, SHEILA SHETH, ROGER C. SANDERS, JONATHAN I. EPSTEIN AND PATRICK C. WALSH From the Departments of Urology, Radiology and Pathology, The Johns Hopkins University School of Medicine, The James Buchanan Brady Urological Institute, The Johns Hopkins Hospital, Baltimore, Maryland
ABSTRACT
To determine the ability of transrectal ultrasound to detect early localized prostate cancer, unsuspected (nonpalpable) cancer in the contralateral lobe of patients undergoing radical prostatectomy for clinically localized disease was evaluated. A total of 59 patients with palpable prostate cancer clinically confined to 1 lobe underwent transrectal ultrasound before radical prostatectomy and step-sectioning of the radical prostatectomy specimen. Transrectal ultrasound was performed with 5 or 7 MHz. real-time transrectal units. Pathological findings in these 59 cases revealed no tumor in the contralateral lobe in 34 (58%) and the presence of unsuspected tumor in 25 (42%). Transrectal ultrasound detected 13 of 25 unsuspected cancers for a sensitivity of 52%. Of 34 patients with no contralateral lobe lesion transrectal ultrasound was correct in 23 for a specificity of 68%. The positive and negative predictive values for transrectal ultrasound in this study group were 54 and 66%, respectively. There was no significant difference in the pathological size of the clinically suspected and clinically unsuspected cancers as measured by average largest dimension, and transrectal ultrasound sensitivity did not correlate with the size of the cancer. Based on careful sonopathological analysis, transrectal ultrasound may not be a good method to detect clinically unsuspected prostate cancer and the false positive rate would appear to be high. (J. Urol., 142: 1008-1010, 1989) Prior studies evaluating the use of transrectal ultrasound in the early detection of prostate cancer often have compared the ability of transrectal ultrasound versus digital rectal examination to detect lesions in asymptomatic patients. 1-6 Because the true prevalence (the denominator) of prostate cancer in these studies is unknown sensitivity and specificity cannot be calculated for transrectal ultrasound. Nevertheless, it has been argued that transrectal ultrasound increases the detection rate of prostate cancer and for this reason some investigators believe its use in screening for early localized prostate cancer is warranted. However, there are no studies that evaluate carefully the ability of transrectal ultrasound to detect unsuspected cancers, since this would require that all prostates screened be examined pathologically, which is impossible. In addition, the transrectal ultrasound diagnosis given before pathological examination would then need to be matched to the pathological diagnosis to correlate sonographic abnormalities with pathological abnormalities. To circumvent these problems and determine the ability of transrectal ultrasound to detect clinically unsuspected prostate cancer we studied a group of patients who underwent transrectal ultrasound and radical prostatectomy with subsequent step-sectioning of the pathological specimen. The correlation between transrectal ultrasound and the subsequent pathological findings forms the basis of this report. METHODS
The records of 59 patients who underwent transrectal ultrasound before radical retropubic prostatectomy were reviewed. All patients had a clinically palpable prostate cancer confined to 1 lobe and pathological confirmation that the palpable cancer did not extend into the contralateral lobe. Prostate specimens were step-sectioned at 3 mm. intervals in a plane perpendicular to the long axis of the prostate, and sections were embedded and examined histologically. The histologically identified prosAccepted for publication April 21, 1989. Supported by Grant Ca 15416 from the National Cancer Institute.
tate cancers were measured in millimeters in 3 dimensions. In the transverse plane of section lesions were measured in the anterior to posterior dimension (up and down) and the transverse dimension (left to right). In addition, the lesions were measured in a cephalocaudad dimension in the sagittal plane by reconstructing the transverse step sections. To identify the location of the lesions the transverse sections were divided into 4 quadrants (left and right anterior and posterior) and a map of the lesion was drawn. Transrectal ultrasound was performed in a transverse and longitudinal plane on the day of or the day before the operation. All radial (transverse) views were obtained with a Bruel and Kjaer model 1846 scanner and a 7.0 MHz. transducer. Longitudinal views were obtained with either a Bruel and Kjaer model 1837 probe and a 7.0 MHz. transducer, or a 5.0 MHz. Aloka SSD-256 linear scanner. Hypoechoic areas in the peripheral zone of the prostate were considered as suspicious for cancer. The location of all lesions suspected of being carcinoma was noted in the transverse and longitudinal planes. In a transverse plane the lesion was indicated as left or right anterior and posterior, and in the longitudinal plane the cephalocaudad location of the lesion was noted. Of the 59 patients 34 (58 %) with a palpable cancer clinically confined to 1 lobe had no cancer in the contralateral lobe on pathological examination. The remaining 25 patients (42%) had a clinically unsuspected cancer in the contralateral lobe in addition to the palpable cancer. All of these unsuspected cancers in the contralateral lobe were in the posterior portion of the prostate or extended partly to the anterior quadrant. In no case did a cancer (palpable or nonpalpable) extend into the contralateral lobe of the prostate on pathological examination. However, in 34 of 59 cases (58%) the palpable cancer demonstrated capsular penetration and/or seminal vesicle invasion on pathological examination. The largest dimension of each cancer was used to calculate the average largest dimension of palpable and nonpalpable cancers. This average largest dimension for palpable and non -
1008
1009
PROSTATE UL'TRASOUNi} FOR D1AG1'l0SIS
palpable disease was with the Mann-Vi/hitney VVilcoxon test for statistical significance. RESULTS
The figure summarizes the transrectal ultrasound findings with respect to the suspected (palpable) and unsuspected (nonpalpable) cancers. Transrectal ultrasound predicted that 11 cancers (false positive results) were present in the 34 cases when there was no contralateral lobe cancer, thus, 23 tests were true negative. Transrectal ultrasound detected 13 (true positive results) of the 25 clinically unsuspected cancers, which means that 12 of the 25 tests were false negative. Therefore, the sensitivity was 52% (true positive tests/true positive tests + false negative tests) and specificity was 68% (true negative tests/true negative tests + false positive tests) for transrectal ultrasound in the detection of unsuspected cancers. Of the 59 suspected palpable cancers transrectal ultrasound detected 54, for a sensitivity of 92%. The table summarizes the sensitivity of transrectal ultrasound with respect to the pathological size of the cancers. The average largest dimension of the 59 palpable cancers was 20.3 mm. compared to 17.2 mm. for the 25 nonpalpable tumors. There was no significant difference in the largest dimension of the tumors between the 2 groups (p = 0.15). The ability of transrectal ultrasound to detect palpable and nonpalpable disease (sensitivity) did not correlate with the size of the lesion as measured in its greatest dimension. Palpable Ca 1 Lobe Involved [n=59)
Non-Palpable Ca 1 Lobe Involved [n=25)
Path Step Section
34
Cancer Detected By TRU 23/34 True Neg=68%
34
11/34 False Pos=32%
NC
25
25
13/25 True Pos=52% 12/25 False Neg=48%
59
54/59 True Pos=92%
X L__________ _ _j
Comparison of preoperative transrectal ultrasound (TRU) findings with pathological fin:lings in 59 step-s~ctioned radical prostatectomy specimens. One lobe mvolved refers to 1act that no chmcally palp~ble or nonpalpable cancer ( Ca) involved contralateral lobe on pathological examination.~, clinically suspected (palpable) cancer. NC, no contralateral lobe cancer. X, clinically unsuspected (nonpalpable) contralateral lobe cancer.
Correlation between pathological size of cancer and transrectal ultrasound sensitivity Pathological Size (mm.)" Suspected Ca (59 pts.):t 6-10 (12) 11-20 (21) 21-30 (20) 31-42 (6) Unsuspected Ca (25 pts.):t 5-10 (6) 11-20 (14) 21-30 (2) 31-45 (3)
Transrectal Ultrasound Sensitivity ( % ) 92 92 90
95 83 52 67 43 50 67
* Numbers in parentheses are numbers of patients. Pathological size is reported as the largest dimension. t Average largest dimension 20.3 mm. t Average largest dimension 17.2 mm.
DiSCUSSIOJ\J
Our study attempts to evaluate the true status of transrectal ultrasound in the detection of clinically unsuspected prostate cancer. Based on this study it appears that transrectal ultrasound may not be as sensitive as reported frequently in the literature in detecting nonpalpable lesions and that the false positive rate is high, For example, Lee and associates recently reported a sensitivity of 91 % for transrectal ultrasound in detecting prostate cancer among 784 self-referred men. 1 Based on the fact that transrectal ultrasound achieved twice the rate of detection than rectal examination it was argued that broader use of transrectal ultrasound in early detection of prostate cancer may be warranted. In that study the specificity was 20%, for a false positive rate of 80% for transrectal ultrasound. However, the true prevalence of disease in this population was unknown. It is possible that the better prostate cancer detection rate for transrectal ultrasound compared to rectal examination is simply a result of an increased number of biopsies that are performed with transrectal ultrasound and not the result of the ability of ultrasound to detect nonpalpable cancer. The largest diameter measured pathologically was used in this study as an indicator of the size of the cancer as seen ultrasound in 1 plane. This measurement is not intended to reflect tumor volume. Since transrectal ultrasound images in 1 plane at a time the largest diameter of the cancer may better reflect trne ultrasound image size as opposed to tumor volume. Surprisingly, the size of the cancer in its largest dimension, whether palpable or nonpalpable, did not correlate with the sensitivity of transrectal ultrasound in detecting the cancer. This also was found by Lee and associates in that the largest cancers were the lesions missed transrectal ultrasound. 1 It may be that other factors, such as the stromal-epithelial ratio of the lesion, can affect the echogenicity of prostate cancer and, thus, the ability of transrectal ultrasound to detect the cancer. Dahnert and associates found that 24 % of 49 prostate cancers scanned in situ after radical prostatectomy were isoechoic. 6 Similarly, Salo and associates found that 30% of 20 prostate cancers scanned in situ after radical prostatectomy were of uniform echogenicity and could not be detected by transrectal ultrasound. 7 These sonographic analyses with pathological correlation suggest that a significant number of cancers are difficult to visualize because of their isoechoic nature. In our study prior biopsy of the contralateral prostatic lobe at initial diagnosis could have affected echogenicity influenced the transrectal ultrasound findings. A scar from a prior biopsy would be expected to influence the specificity and have little influence on the sensitivity of trnnsrectal ultrasound in detecting nonpalpable cancer. Nevertheless, careful evaluation of the factors that influence the eciio11e111c1~,, of cancer ological correlation will be important to determine if transrectal ultrasound truly is a useful test for the detection of prostatic cancer. The positive and negative predictive values of a test often are calculated to evaluate the test for use in screening for a disease. The value of these calculations is that even though the sensitivity and specificity of a test are relatively high, if the prevalence of the disease in the population being screened is low the test may not be valuable for use in screening. The predictive value of a positive test is the true positive tests divided by the total number of positive tests (false and true positive tests), and indicates how many times the disease being screened for actually will be present if the test is positive. In this study population with a prevalence of unsuspected cancer of 42% (25 of 59) the positive predictive value can be calculated with the aforementioned definition to be 54%. The predictive value of a negative test is the true negative tests divided by the total number of negative tests (true negative and false negative). This indicates how many times the disease being screened for actually is excluded if the test is negative. In the study population it can be calculated that the negative predictive
1010
CARTER AND ASSOCIATES
value is 66%. Given the same sensitivity and specificity, as the prevalence of the disease decreases in the population being screened the positive predictive value of the test will decrease and the negative predictive value will increase. This is .true because with less cases of the disease in the population the true positive tests (numerator of positive predictive value) decrease dramatically relative to the false positive tests, and the true negative tests (numerator of negative predictive value) increase dramatically relative to the false negative tests. Prior studies have found a positive predictive value of approximately 15 to 30% for transrectal ultrasound. 1- 3 However, it should be emphasized that the true prevalence in these studies was not known and, therefore, the predictive value is approximated. With the sensitivity and specificity of transrectal ultrasound found in this study population, if the prevalence of unsuspected clinical prostate cancer in a screened population was 10% the predictive value of a positive test would be 15% and the predictive value of a negative test would be 92%. Thus, in this setting only 15% of the men with an abnormal transrectal ultrasound would actually have cancer. If transrectal ultrasound were negative prostate cancer would be excluded correctly 92% of the time. On the other hand, if these calculations are based on the prevalence of histological prostate cancers commonly found at autopsy in a screened population (that is 30%) the positive and negative predictive values would be 42 and 77%, respectively. Therefore, the use of transrectal ultrasound to screen for early, localized prostate cancer would appear to be prema-
ture and will require further careful evaluation of the true ability of this modality to detect clinically unsuspected cancer. REFERENCES 1. Lee, F., Littrup, P. J., Torp-Pedersen, S. T., Mettlin, C., McHugh,
2. 3.
4.
5. 6.
7.
T. A., Gray, J. M., Kumasaka, G. H. and McLeary, R. D.: Prostate cancer: comparison of transrectal US and digital rectal examination for screening. Radiology, 168: 389, 1988. Rifkin, M. D. and Choi, H.: Implications of small, peripheral hypoechoic lesions in endorectal US of the prostate. Radiology, 166: 619, 1988. Cooner, W. H., Mosley, B. R., Rutherford, C. L., Jr., Beard, J. H., Pond, H. S., Bass, R. B., Jr. and Terry, W. J.: Clinical application of transrectal ultrasonography and prostate specific antigen in the search for prostate cancer. J. Urol., 139: 758, 1988. Rosenberg, S., Sogani, P. C., Parmer, E. A. and Miller, D. G.: Screening of ambulatory patients for prostate cancer by transrectal ultrasonography. J. Urol., part 2, 137: 241A, abstract 551, 1987. Watanabe, H., Ohe, H., Inabe, T., Itakura, Y., Saitoh, M. and Nakao, M.: A mobile mass screening unit for prostatic disease. Prostate, 5: 559, 1984. Dahnert, W. F., Hamper, U. M., Eggleston, J.C., Walsh, P. C. and Sanders, R. C.: Prostatic evaluation by transrectal sonography with histopathologic correlation: the echopenic appearance of early carcinoma. Radiology, 158: 97, 1986. Salo, J. 0., Rannikko, S., Miikinen, J. and Lehtonen, T.: Echogenic structure of prostatic cancer imaged on radical prostatectomy specimens. Prostate, 10: 1, 1987.
THE JOURNAL OF UROLOGY Copyright© 1989 by AMERICAN UROLOGICAL ASSOCIATION, INC.
Printed in U.S.A.
EVALUATION OF TRANSRECTAL ULTRASOUND IN THE EARLY DETECTION OF PROSTATE CANCER H. BALLENTINE CARTER, ULRIKE M. HAMPER, SHEILA SHETH, ROGER C. SANDERS, JONATHAN I. EPSTEIN AND PATRICK C. WALSH From the Departments of Urology, Radiology and Pathology, The Johns Hopkins University School of Medicine, The James Buchanan Brady Urological Institute, The Johns Hopkins Hospital, Baltimore, Maryland
ABSTRACT
To determine the ability of transrectal ultrasound to detect early localized prostate cancer, unsuspected (nonpalpable) cancer in the contralateral lobe of patients undergoing radical prostatectomy for clinically localized disease was evaluated. A total of 59 patients with palpable prostate cancer clinically confined to 1 lobe underwent transrectal ultrasound before radical prostatectomy and step-sectioning of the radical prostatectomy specimen. Transrectal ultrasound was performed with 5 or 7 MHz. real-time transrectal units. Pathological findings in these 59 cases revealed no tumor in the contralateral lobe in 34 (58%) and the presence of unsuspected tumor in 25 (42%). Transrectal ultrasound detected 13 of 25 unsuspected cancers for a sensitivity of 52%. Of 34 patients with no contralateral lobe lesion transrectal ultrasound was correct in 23 for a specificity of 68%. The positive and negative predictive values for transrectal ultrasound in this study group were 54 and 66%, respectively. There was no significant difference in the pathological size of the clinically suspected and clinically unsuspected cancers as measured by average largest dimension, and transrectal ultrasound sensitivity did not correlate with the size of the cancer. Based on careful sonopathological analysis, transrectal ultrasound may not be a good method to detect clinically unsuspected prostate cancer and the false positive rate would appear to be high. (J. Urol., 142: 1008-1010, 1989) Prior studies evaluating the use of transrectal ultrasound in the early detection of prostate cancer often have compared the ability of transrectal ultrasound versus digital rectal examination to detect lesions in asymptomatic patients. 1-6 Because the true prevalence (the denominator) of prostate cancer in these studies is unknown sensitivity and specificity cannot be calculated for transrectal ultrasound. Nevertheless, it has been argued that transrectal ultrasound increases the detection rate of prostate cancer and for this reason some investigators believe its use in screening for early localized prostate cancer is warranted. However, there are no studies that evaluate carefully the ability of transrectal ultrasound to detect unsuspected cancers, since this would require that all prostates screened be examined pathologically, which is impossible. In addition, the transrectal ultrasound diagnosis given before pathological examination would then need to be matched to the pathological diagnosis to correlate sonographic abnormalities with pathological abnormalities. To circumvent these problems and determine the ability of transrectal ultrasound to detect clinically unsuspected prostate cancer we studied a group of patients who underwent transrectal ultrasound and radical prostatectomy with subsequent step-sectioning of the pathological specimen. The correlation between transrectal ultrasound and the subsequent pathological findings forms the basis of this report. METHODS
The records of 59 patients who underwent transrectal ultrasound before radical retropubic prostatectomy were reviewed. All patients had a clinically palpable prostate cancer confined to 1 lobe and pathological confirmation that the palpable cancer did not extend into the contralateral lobe. Prostate specimens were step-sectioned at 3 mm. intervals in a plane perpendicular to the long axis of the prostate, and sections were embedded and examined histologically. The histologically identified prosAccepted for publication April 21, 1989. Supported by Grant Ca 15416 from the National Cancer Institute.
tate cancers were measured in millimeters in 3 dimensions. In the transverse plane of section lesions were measured in the anterior to posterior dimension (up and down) and the transverse dimension (left to right). In addition, the lesions were measured in a cephalocaudad dimension in the sagittal plane by reconstructing the transverse step sections. To identify the location of the lesions the transverse sections were divided into 4 quadrants (left and right anterior and posterior) and a map of the lesion was drawn. Transrectal ultrasound was performed in a transverse and longitudinal plane on the day of or the day before the operation. All radial (transverse) views were obtained with a Bruel and Kjaer model 1846 scanner and a 7.0 MHz. transducer. Longitudinal views were obtained with either a Bruel and Kjaer model 1837 probe and a 7.0 MHz. transducer, or a 5.0 MHz. Aloka SSD-256 linear scanner. Hypoechoic areas in the peripheral zone of the prostate were considered as suspicious for cancer. The location of all lesions suspected of being carcinoma was noted in the transverse and longitudinal planes. In a transverse plane the lesion was indicated as left or right anterior and posterior, and in the longitudinal plane the cephalocaudad location of the lesion was noted. Of the 59 patients 34 (58 %) with a palpable cancer clinically confined to 1 lobe had no cancer in the contralateral lobe on pathological examination. The remaining 25 patients (42%) had a clinically unsuspected cancer in the contralateral lobe in addition to the palpable cancer. All of these unsuspected cancers in the contralateral lobe were in the posterior portion of the prostate or extended partly to the anterior quadrant. In no case did a cancer (palpable or nonpalpable) extend into the contralateral lobe of the prostate on pathological examination. However, in 34 of 59 cases (58%) the palpable cancer demonstrated capsular penetration and/or seminal vesicle invasion on pathological examination. The largest dimension of each cancer was used to calculate the average largest dimension of palpable and nonpalpable cancers. This average largest dimension for palpable and non -
1008
1009
PROSTATE UL'TRASOUNi} FOR D1AG1'l0SIS
palpable disease was with the Mann-Vi/hitney VVilcoxon test for statistical significance. RESULTS
The figure summarizes the transrectal ultrasound findings with respect to the suspected (palpable) and unsuspected (nonpalpable) cancers. Transrectal ultrasound predicted that 11 cancers (false positive results) were present in the 34 cases when there was no contralateral lobe cancer, thus, 23 tests were true negative. Transrectal ultrasound detected 13 (true positive results) of the 25 clinically unsuspected cancers, which means that 12 of the 25 tests were false negative. Therefore, the sensitivity was 52% (true positive tests/true positive tests + false negative tests) and specificity was 68% (true negative tests/true negative tests + false positive tests) for transrectal ultrasound in the detection of unsuspected cancers. Of the 59 suspected palpable cancers transrectal ultrasound detected 54, for a sensitivity of 92%. The table summarizes the sensitivity of transrectal ultrasound with respect to the pathological size of the cancers. The average largest dimension of the 59 palpable cancers was 20.3 mm. compared to 17.2 mm. for the 25 nonpalpable tumors. There was no significant difference in the largest dimension of the tumors between the 2 groups (p = 0.15). The ability of transrectal ultrasound to detect palpable and nonpalpable disease (sensitivity) did not correlate with the size of the lesion as measured in its greatest dimension. Palpable Ca 1 Lobe Involved [n=59)
Non-Palpable Ca 1 Lobe Involved [n=25)
Path Step Section
34
Cancer Detected By TRU 23/34 True Neg=68%
34
11/34 False Pos=32%
NC
25
25
13/25 True Pos=52% 12/25 False Neg=48%
59
54/59 True Pos=92%
X L__________ _ _j
Comparison of preoperative transrectal ultrasound (TRU) findings with pathological fin:lings in 59 step-s~ctioned radical prostatectomy specimens. One lobe mvolved refers to 1act that no chmcally palp~ble or nonpalpable cancer ( Ca) involved contralateral lobe on pathological examination.~, clinically suspected (palpable) cancer. NC, no contralateral lobe cancer. X, clinically unsuspected (nonpalpable) contralateral lobe cancer.
Correlation between pathological size of cancer and transrectal ultrasound sensitivity Pathological Size (mm.)" Suspected Ca (59 pts.):t 6-10 (12) 11-20 (21) 21-30 (20) 31-42 (6) Unsuspected Ca (25 pts.):t 5-10 (6) 11-20 (14) 21-30 (2) 31-45 (3)
Transrectal Ultrasound Sensitivity ( % ) 92 92 90
95 83 52 67 43 50 67
* Numbers in parentheses are numbers of patients. Pathological size is reported as the largest dimension. t Average largest dimension 20.3 mm. t Average largest dimension 17.2 mm.
DiSCUSSIOJ\J
Our study attempts to evaluate the true status of transrectal ultrasound in the detection of clinically unsuspected prostate cancer. Based on this study it appears that transrectal ultrasound may not be as sensitive as reported frequently in the literature in detecting nonpalpable lesions and that the false positive rate is high, For example, Lee and associates recently reported a sensitivity of 91 % for transrectal ultrasound in detecting prostate cancer among 784 self-referred men. 1 Based on the fact that transrectal ultrasound achieved twice the rate of detection than rectal examination it was argued that broader use of transrectal ultrasound in early detection of prostate cancer may be warranted. In that study the specificity was 20%, for a false positive rate of 80% for transrectal ultrasound. However, the true prevalence of disease in this population was unknown. It is possible that the better prostate cancer detection rate for transrectal ultrasound compared to rectal examination is simply a result of an increased number of biopsies that are performed with transrectal ultrasound and not the result of the ability of ultrasound to detect nonpalpable cancer. The largest diameter measured pathologically was used in this study as an indicator of the size of the cancer as seen ultrasound in 1 plane. This measurement is not intended to reflect tumor volume. Since transrectal ultrasound images in 1 plane at a time the largest diameter of the cancer may better reflect trne ultrasound image size as opposed to tumor volume. Surprisingly, the size of the cancer in its largest dimension, whether palpable or nonpalpable, did not correlate with the sensitivity of transrectal ultrasound in detecting the cancer. This also was found by Lee and associates in that the largest cancers were the lesions missed transrectal ultrasound. 1 It may be that other factors, such as the stromal-epithelial ratio of the lesion, can affect the echogenicity of prostate cancer and, thus, the ability of transrectal ultrasound to detect the cancer. Dahnert and associates found that 24 % of 49 prostate cancers scanned in situ after radical prostatectomy were isoechoic. 6 Similarly, Salo and associates found that 30% of 20 prostate cancers scanned in situ after radical prostatectomy were of uniform echogenicity and could not be detected by transrectal ultrasound. 7 These sonographic analyses with pathological correlation suggest that a significant number of cancers are difficult to visualize because of their isoechoic nature. In our study prior biopsy of the contralateral prostatic lobe at initial diagnosis could have affected echogenicity influenced the transrectal ultrasound findings. A scar from a prior biopsy would be expected to influence the specificity and have little influence on the sensitivity of trnnsrectal ultrasound in detecting nonpalpable cancer. Nevertheless, careful evaluation of the factors that influence the eciio11e111c1~,, of cancer ological correlation will be important to determine if transrectal ultrasound truly is a useful test for the detection of prostatic cancer. The positive and negative predictive values of a test often are calculated to evaluate the test for use in screening for a disease. The value of these calculations is that even though the sensitivity and specificity of a test are relatively high, if the prevalence of the disease in the population being screened is low the test may not be valuable for use in screening. The predictive value of a positive test is the true positive tests divided by the total number of positive tests (false and true positive tests), and indicates how many times the disease being screened for actually will be present if the test is positive. In this study population with a prevalence of unsuspected cancer of 42% (25 of 59) the positive predictive value can be calculated with the aforementioned definition to be 54%. The predictive value of a negative test is the true negative tests divided by the total number of negative tests (true negative and false negative). This indicates how many times the disease being screened for actually is excluded if the test is negative. In the study population it can be calculated that the negative predictive
1010
CARTER AND ASSOCIATES
value is 66%. Given the same sensitivity and specificity, as the prevalence of the disease decreases in the population being screened the positive predictive value of the test will decrease and the negative predictive value will increase. This is .true because with less cases of the disease in the population the true positive tests (numerator of positive predictive value) decrease dramatically relative to the false positive tests, and the true negative tests (numerator of negative predictive value) increase dramatically relative to the false negative tests. Prior studies have found a positive predictive value of approximately 15 to 30% for transrectal ultrasound. 1- 3 However, it should be emphasized that the true prevalence in these studies was not known and, therefore, the predictive value is approximated. With the sensitivity and specificity of transrectal ultrasound found in this study population, if the prevalence of unsuspected clinical prostate cancer in a screened population was 10% the predictive value of a positive test would be 15% and the predictive value of a negative test would be 92%. Thus, in this setting only 15% of the men with an abnormal transrectal ultrasound would actually have cancer. If transrectal ultrasound were negative prostate cancer would be excluded correctly 92% of the time. On the other hand, if these calculations are based on the prevalence of histological prostate cancers commonly found at autopsy in a screened population (that is 30%) the positive and negative predictive values would be 42 and 77%, respectively. Therefore, the use of transrectal ultrasound to screen for early, localized prostate cancer would appear to be prema-
ture and will require further careful evaluation of the true ability of this modality to detect clinically unsuspected cancer. REFERENCES 1. Lee, F., Littrup, P. J., Torp-Pedersen, S. T., Mettlin, C., McHugh,
2. 3.
4.
5. 6.
7.
T. A., Gray, J. M., Kumasaka, G. H. and McLeary, R. D.: Prostate cancer: comparison of transrectal US and digital rectal examination for screening. Radiology, 168: 389, 1988. Rifkin, M. D. and Choi, H.: Implications of small, peripheral hypoechoic lesions in endorectal US of the prostate. Radiology, 166: 619, 1988. Cooner, W. H., Mosley, B. R., Rutherford, C. L., Jr., Beard, J. H., Pond, H. S., Bass, R. B., Jr. and Terry, W. J.: Clinical application of transrectal ultrasonography and prostate specific antigen in the search for prostate cancer. J. Urol., 139: 758, 1988. Rosenberg, S., Sogani, P. C., Parmer, E. A. and Miller, D. G.: Screening of ambulatory patients for prostate cancer by transrectal ultrasonography. J. Urol., part 2, 137: 241A, abstract 551, 1987. Watanabe, H., Ohe, H., Inabe, T., Itakura, Y., Saitoh, M. and Nakao, M.: A mobile mass screening unit for prostatic disease. Prostate, 5: 559, 1984. Dahnert, W. F., Hamper, U. M., Eggleston, J.C., Walsh, P. C. and Sanders, R. C.: Prostatic evaluation by transrectal sonography with histopathologic correlation: the echopenic appearance of early carcinoma. Radiology, 158: 97, 1986. Salo, J. 0., Rannikko, S., Miikinen, J. and Lehtonen, T.: Echogenic structure of prostatic cancer imaged on radical prostatectomy specimens. Prostate, 10: 1, 1987.