The Prostate Supplement 4: 51-57(1992)
Volume of Normal Prostate, of Prostate Cancer, and of Benign Prostatic Hyperplasia: Are Correlations With Prostate Specific Antigen Clinically Useful? R. Clements, M.D. Penney, R.J. Etherington, G.J. Griffiths, H. Hughes, and
W.B.Peeling Departments of Radiology and Urology, St. Woolos Hospital, Newport (R.C., R.J.E., G.J. G., W.B.P.), and Department of Chemical Pathology, Royal Gwent Hospital, Newport (M.0 .P., H.H.j, United Kingdom
This retrospective study correlated prostate volume, determined by transrectal ultrasonography, with s m prostate specific antigen (PSA), by k i n g regression analysis, in patients with confirmed benign prostatic hyperplasia (BPH) and patients with non-metastatic (MO) or metastatic (Ml) prostate cancer. In BPH, a highly significant correlationwas found between log,,[PSA] and prostate volume. When this PSAholume regressionpattern for BPH was used asa referencestandard, all 17patients with M1 prostate cancer and 83% of the 23 patients with MO disease were discriminated from BPH. Key words: Deming regression analysis, prostate cancer diagnosis, transrectal ultrasonography
INTRODUCTION
Measurement of prostate volume, usually by transrectal ultrasound (TRUS), and assay of prostate specific antigen (PSA) are now routine procedures in most urologic departments. Although PSA is now acceptedto be a more sensitive serum marker for prostate cancerthan prostatic acidphosphatase[11there has always been uncertainty about the application of PSA findings to the diagnosis of prostatic disorders, particularly cancer. PSA is unique in that it is produced solely by epithelial cells of the prostate [2,3], and serum levels of PSA reflect the overall epithelial cell mass within the gland. Therefore, tissue concentrations of PSA shouldtheoreticallybe greaterin prostate cancerthan inbenignprostatichyperplasia (BPH) and this should be reflected in serum levels of PSA. This concept is
Address correspondence to Professor W.B. Peeling, Department of Urology, St. Woolos Hospital, 131 Stow Hill, Newport, Gwent, UK.
0 1992 Wiley-Liss, Inc.
52
Clements et al.
supported by the results of Stamey et al., who showed that in BPH, serum PSA levels were elevated by approximately 0.3 1ng/ml for each gram of BPH tissue [4], but by 3.5 ng/ml for each gram of prostate cancer [5]. As the specific gravity of prostate is nearly unity [ 5 ] , the prostatic weight in grams should be equivalent to the prostatic volume in ml. In Newport the relationship between prostate volume and serum PSA has been studied inpatients with histologically proven BPH orprostatic cancer in order to evaluate the usefulness of volume-corrected PSA for distinguishing patients with BPH from those with cancer. MATERIALS AND METHODS
PSAmeasurements and prostate volumes oftwo groups ofpatients have been analyzed retrospectively. In group 1, there were 50 patients in whom the prostate was considered benign by both digital rectal examination and transrectal ultrasound (TRUS). Subsequent transurethral resection of the prostate (TURP) with histological examination of the chippings demonstrated BPH but no evidence of malignancy. In group 2,40 patients with histologically proven prostate cancer, in whom the diagnosis had been made by either prostatic biopsy or by TURP, were studied. Subdivision of this group into those without apparent metastatic disease (MO; 23 patients) and those with proven metastases (Ml; 17 patients) followed assessment by chest radiographs and radioisotope bone scans. In all patients, PSA and prostatic volume measurements were obtained within 4 weeks of each other. TRUS examinations were undertaken with a Bruel & Kjaer 1846 scanner with either a 4MHz radial transducer or a 7MHz multiplanar transducer. All volume measurements were obtained by step section planimetry following cross sectional measurements of the area of the prostate at 0.5 cm intervals. PSA measurements were carried out using a two site immuno-radiometric assay (Hybritech Inc.). The relationship between serum PSA and prostatic volume in BPH was analyzed by Deming regression analysis using the equation: log,,, PSA = 5.62 x (UPV) + 0.32 r = 0.62 standard deviation of residual error of regression about Y = 0.30 where PSA = serum prostatic specific antigen (ng/ml) UPV = TRUS volume (ml). RESULTS
Patients with BPH (n=50) were aged from 53 to 86 years (mean 72 years). Their prostatic volumes varied from 8 to 244 ml (mean 62 ml) and their serum PSA levels varied from 1.3 to 85 ng/ml (mean 7.7 ng/ml).
Correlations of Prostate Volume and PSA
53
Patients withMO prostatic cancer (n=23) were aged from 50 to 90 years (mean 73 years). Prostatic volumes varied from 9 to 151ml (mean49 ml) and their serum PSA levels ranged from 3.8 to 722 ng/ml (mean 95 ng/ml). Patients with established M1 disease (n=17) were aged from 53 to 83 years (mean 65 years). Prostatic volumes in these patients varied from 17 to 62 ml (mean 35 ml) and serum PSA levels ranged from 47 to 6890 ng/ml (mean 1337 ng/ml). Analysis
Comparison of serum PSA levels (Fig. 1) showed non-parametric distribution within groups of patients with BPH, MO or M1 prostate cancer. Group comparisons using Wilcoxon Rank Sum analysis showed highly significant differences of distribution of PSA between patients with BPH and MO disease (P < 0.001) and patients with BPH and M1 disease (P< 0.001). There was a highly significant correlation between serum PSA and prostate volume in patients with BPH with the best correlation between log-transformed PSA and non-transformed prostate volumes (Fig. 2). All PSA values except one fell within the 97.5% confidence limits ofthe regression for patients with BPH. In 9patients(18%),PSAvaluesweregreaterthan lOng/ml,but 8 ofthese(l6%) were within the 97.5% confidence limits. Seven ofthese patients had prostate volumes exceeding 50 ml. In other words, 41/50 (82%) ofPSA values inpatients with BPH were less than 10 ng/ml, irrespective of the prostate volume. Of 23 men with MO prostate cancer, PSA levels in 4 (17%) were less than 10 ng/l (Fig. 3). Of the remaining 19/23 (83%), serum PSA levels were greater than the 97.5% confidence limits of the BPH group. Of 17 men with M1 prostatic cancer, serum PSA levels were all well beyond the 97.5% confidence limits of patients with BPH (Fig. 4).
= 1000
. E
0
c
Y
3 a
t
1
loot
t 11
I
BPH
I
MO
I
M1 Stage
Fig. 1. Serum prostate specific antigen (PSA) in benign prostatic hyperplasia (BPH) and prostate cancer. Wilcoxon Rank Sum analysis showed highly significant differences ( P < 0.001) between BPH and MO disease and between BPH and M1.
Clements et al.
54
97.5% confidence limit
2-
Mean
100
50
2
150
200
250
Prostate volume (ml)
patients with BPH Mean for patients with BPH
0'
I
I
I
1
I
50
100
150
200
250
3 =
E .3 0
c
v
a
Prostate volume (ml)
... . 2
0
.
97.5% confidence limit for
Correlations of Prostate Volume and PSA
55
DISCUSSION
The problem with PSA values has always been the overlap that may occur between people with normal prostates, patients with BPH, and patients with cancer -- particularly at low assay values. In our study, though there were significant differences in distribution of serum PSA between patients with BPH and MO disease, there was some overlap between the two groups. There was also some overlap between PSA in patients with BPH and M1 cancer, so the usefblness of s e w n PSA levels to separate patients with BPH from those with cancer was very limited (Fig. 1.) A different pattern was evident when prostate volume was introduced as a variable against PSA with submission to Deming regression analysis. In patients with BPH, volume correction using the Deming regression equation produced a particular pattern of PSA that was within the 97.5% confidence limits (Fig. 2). When the PSNvolume regression pattern of patients with BPH was used as a reference standard for patients with prostate cancer, it was evident that all patients with M l cancer were clearly discriminatedaway from the BPHpattern (Fig. 4) and that 83% of patients with MO cancers were similarly discriminated (Fig. 3). Only 4 (17%) of MO patients were not discriminated from the BPH pattern, and all of these had prostate volumes of less than 50 ml. CONCLUSIONS
An encouraging discriminationwas achieved between patients with BPH and prostate cancer. In this preliminary but retrospective study using Deming regression analysis to compare log-transformed PSA with prostatic volume, all patients with MI disease and 83% of those with MO disease were discriminated from those with BPH. A prospective study to examine in depth the concept of volume adjustment of PSA is being started. The idea could be practically applied as a volume-correctedchart for BPHusedas atemplateand kept forreference inclinics, wards, and scanning rooms, so that when a patient undergoes TRUS volume measurement, a quick comparison of the PSA in relation to the 97.5% confidence limits for that particular volume may give some indication of the significance of the PSA result. Alternatively,the software ofthe scanner could be programmed to indicate this information.
Fig. 2. Correlationbetween prostatevolume and serum prostate specific antigen (PSA) inpatients with benign prostatic hyperplasia. Fig. 3. Prostate volume and serum PSA in patients with non-metastatic prostate cancer in comparison with patients with benign prostatic hyperplasia (BPH). Fig. 4. Prostate volume and serum prostate specific antigen (PSA) in patients with metastatic prostate cancer in comparison with patients with benign prostatic hyperplasia (BPH).
56
Clements et al.
Is correlation between prostate volume and PSA clinically useful? Deming regression analysis of PSA and prostate volume is certainly very promising as a sorting process between patients with BPH and cancer, but its application using normal prostates as reference standards is not yet clear. REFERENCES
1. Oosterom R, Bogdanowicz J, Schrader FH: Evaluation of prostate-specificantigen in untreated prostatic carcinoma. Em Urn1 16253-257,1989. 2. Nadji M, Tabei SZ, Castro A, Chu TM, Murphy GP, Wang MC, Morales AR: Prostatic specificantigen:animmuno-histologicmarkerforprostaticneoplasms.Cancer48:12291232, 1981. 3. Frankel AE, Rouse RV, Wang MC, Chu TM, HerzenbergL A Monoclonal antibodies to a human prostatic antigen. Cancer Res 42:37143718,1982. 4. Stamey TA, Kabalin JN: Prostate specific antigen in the diagnosis and treatment of adenocarcinomaof the prostate I. Untreated patients. J Urol 141:1070-1075,1989. 5. Watanabe H, Igari D, Tanahashi Y, Harada K, Saitoh M: Measurementsof size and weight of prostate by means of transrectal ultrmnotomography. Tohoku J Exp Med 114~227-285,1974.
QUESTIONS
S. Fossil (Oslo, Norway): Professor Peeling, did you say that you had evidence of increased PSA in every M+ disease patient? I have seen M+ patients with a normal PSA. W.B. Peeling: These results were from a group of patients who were studied retrospectively. We now feel encouraged to do a full study and are looking at the idea of a volume-corrected PSA. J.E. Montie (Fort Lauderdale, FL): Inpatients withbenignprostatichyperplasia, there is an unknown amount of cancer present that could contribute to the PSA elevation. How will it be possible to account for that? W.B. Peeling: I don’t think it can be accounted for in a clinical study. Most of the patients on the chart, however, had prostatic volumes of 40 ml or more. In a relatively low volume prostate, which is normal on TRUS, we would be suspicious if PSA was elevated above, for example, 20 ng/ml. We look upon PSA measurement as a means of identifying patients who probably have prostate cancer, and the area of particular interest is 4-10 ng/ml. S.Fossil: Do you thinkthat PSA really discriminates patients with pathological M+ disease? Many of the patients with advanced MO disease will have lymph node involvement, thus confusing the interpretation of the results in this group. W.B. Peeling: It is hoped that diagnostic processes using TRUS and MRI will improve the sensitivity of imaging and also that the PSA assay will improve with
Correlations of Prostate Volume and PSA
57
time. At the moment, we are working with a normal range of 0-4ng/ml (Hybritech). If the sensitivity of the process was increased so that the normal range was 0-100, the technique may become more accurate. F.K.Mostofi (Washington, D.C.):Not every prostate cancer produces PSA. There are cancers of the prostate, particularly poorly differentiated cancers, which irrespective of their volume do not produce PSA.
Volume of Normal Prostate, of Prostate Cancer, and of Benign Prostatic Hyperplasia: Are Correlations With Prostate Specific Antigen Clinically Useful? R. Clements, M.D. Penney, R.J. Etherington, G.J. Griffiths, H. Hughes, and
W.B.Peeling Departments of Radiology and Urology, St. Woolos Hospital, Newport (R.C., R.J.E., G.J. G., W.B.P.), and Department of Chemical Pathology, Royal Gwent Hospital, Newport (M.0 .P., H.H.j, United Kingdom
This retrospective study correlated prostate volume, determined by transrectal ultrasonography, with s m prostate specific antigen (PSA), by k i n g regression analysis, in patients with confirmed benign prostatic hyperplasia (BPH) and patients with non-metastatic (MO) or metastatic (Ml) prostate cancer. In BPH, a highly significant correlationwas found between log,,[PSA] and prostate volume. When this PSAholume regressionpattern for BPH was used asa referencestandard, all 17patients with M1 prostate cancer and 83% of the 23 patients with MO disease were discriminated from BPH. Key words: Deming regression analysis, prostate cancer diagnosis, transrectal ultrasonography
INTRODUCTION
Measurement of prostate volume, usually by transrectal ultrasound (TRUS), and assay of prostate specific antigen (PSA) are now routine procedures in most urologic departments. Although PSA is now acceptedto be a more sensitive serum marker for prostate cancerthan prostatic acidphosphatase[11there has always been uncertainty about the application of PSA findings to the diagnosis of prostatic disorders, particularly cancer. PSA is unique in that it is produced solely by epithelial cells of the prostate [2,3], and serum levels of PSA reflect the overall epithelial cell mass within the gland. Therefore, tissue concentrations of PSA shouldtheoreticallybe greaterin prostate cancerthan inbenignprostatichyperplasia (BPH) and this should be reflected in serum levels of PSA. This concept is
Address correspondence to Professor W.B. Peeling, Department of Urology, St. Woolos Hospital, 131 Stow Hill, Newport, Gwent, UK.
0 1992 Wiley-Liss, Inc.
52
Clements et al.
supported by the results of Stamey et al., who showed that in BPH, serum PSA levels were elevated by approximately 0.3 1ng/ml for each gram of BPH tissue [4], but by 3.5 ng/ml for each gram of prostate cancer [5]. As the specific gravity of prostate is nearly unity [ 5 ] , the prostatic weight in grams should be equivalent to the prostatic volume in ml. In Newport the relationship between prostate volume and serum PSA has been studied inpatients with histologically proven BPH orprostatic cancer in order to evaluate the usefulness of volume-corrected PSA for distinguishing patients with BPH from those with cancer. MATERIALS AND METHODS
PSAmeasurements and prostate volumes oftwo groups ofpatients have been analyzed retrospectively. In group 1, there were 50 patients in whom the prostate was considered benign by both digital rectal examination and transrectal ultrasound (TRUS). Subsequent transurethral resection of the prostate (TURP) with histological examination of the chippings demonstrated BPH but no evidence of malignancy. In group 2,40 patients with histologically proven prostate cancer, in whom the diagnosis had been made by either prostatic biopsy or by TURP, were studied. Subdivision of this group into those without apparent metastatic disease (MO; 23 patients) and those with proven metastases (Ml; 17 patients) followed assessment by chest radiographs and radioisotope bone scans. In all patients, PSA and prostatic volume measurements were obtained within 4 weeks of each other. TRUS examinations were undertaken with a Bruel & Kjaer 1846 scanner with either a 4MHz radial transducer or a 7MHz multiplanar transducer. All volume measurements were obtained by step section planimetry following cross sectional measurements of the area of the prostate at 0.5 cm intervals. PSA measurements were carried out using a two site immuno-radiometric assay (Hybritech Inc.). The relationship between serum PSA and prostatic volume in BPH was analyzed by Deming regression analysis using the equation: log,,, PSA = 5.62 x (UPV) + 0.32 r = 0.62 standard deviation of residual error of regression about Y = 0.30 where PSA = serum prostatic specific antigen (ng/ml) UPV = TRUS volume (ml). RESULTS
Patients with BPH (n=50) were aged from 53 to 86 years (mean 72 years). Their prostatic volumes varied from 8 to 244 ml (mean 62 ml) and their serum PSA levels varied from 1.3 to 85 ng/ml (mean 7.7 ng/ml).
Correlations of Prostate Volume and PSA
53
Patients withMO prostatic cancer (n=23) were aged from 50 to 90 years (mean 73 years). Prostatic volumes varied from 9 to 151ml (mean49 ml) and their serum PSA levels ranged from 3.8 to 722 ng/ml (mean 95 ng/ml). Patients with established M1 disease (n=17) were aged from 53 to 83 years (mean 65 years). Prostatic volumes in these patients varied from 17 to 62 ml (mean 35 ml) and serum PSA levels ranged from 47 to 6890 ng/ml (mean 1337 ng/ml). Analysis
Comparison of serum PSA levels (Fig. 1) showed non-parametric distribution within groups of patients with BPH, MO or M1 prostate cancer. Group comparisons using Wilcoxon Rank Sum analysis showed highly significant differences of distribution of PSA between patients with BPH and MO disease (P < 0.001) and patients with BPH and M1 disease (P< 0.001). There was a highly significant correlation between serum PSA and prostate volume in patients with BPH with the best correlation between log-transformed PSA and non-transformed prostate volumes (Fig. 2). All PSA values except one fell within the 97.5% confidence limits ofthe regression for patients with BPH. In 9patients(18%),PSAvaluesweregreaterthan lOng/ml,but 8 ofthese(l6%) were within the 97.5% confidence limits. Seven ofthese patients had prostate volumes exceeding 50 ml. In other words, 41/50 (82%) ofPSA values inpatients with BPH were less than 10 ng/ml, irrespective of the prostate volume. Of 23 men with MO prostate cancer, PSA levels in 4 (17%) were less than 10 ng/l (Fig. 3). Of the remaining 19/23 (83%), serum PSA levels were greater than the 97.5% confidence limits of the BPH group. Of 17 men with M1 prostatic cancer, serum PSA levels were all well beyond the 97.5% confidence limits of patients with BPH (Fig. 4).
= 1000
. E
0
c
Y
3 a
t
1
loot
t 11
I
BPH
I
MO
I
M1 Stage
Fig. 1. Serum prostate specific antigen (PSA) in benign prostatic hyperplasia (BPH) and prostate cancer. Wilcoxon Rank Sum analysis showed highly significant differences ( P < 0.001) between BPH and MO disease and between BPH and M1.
Clements et al.
54
97.5% confidence limit
2-
Mean
100
50
2
150
200
250
Prostate volume (ml)
patients with BPH Mean for patients with BPH
0'
I
I
I
1
I
50
100
150
200
250
3 =
E .3 0
c
v
a
Prostate volume (ml)
... . 2
0
.
97.5% confidence limit for
Correlations of Prostate Volume and PSA
55
DISCUSSION
The problem with PSA values has always been the overlap that may occur between people with normal prostates, patients with BPH, and patients with cancer -- particularly at low assay values. In our study, though there were significant differences in distribution of serum PSA between patients with BPH and MO disease, there was some overlap between the two groups. There was also some overlap between PSA in patients with BPH and M1 cancer, so the usefblness of s e w n PSA levels to separate patients with BPH from those with cancer was very limited (Fig. 1.) A different pattern was evident when prostate volume was introduced as a variable against PSA with submission to Deming regression analysis. In patients with BPH, volume correction using the Deming regression equation produced a particular pattern of PSA that was within the 97.5% confidence limits (Fig. 2). When the PSNvolume regression pattern of patients with BPH was used as a reference standard for patients with prostate cancer, it was evident that all patients with M l cancer were clearly discriminatedaway from the BPHpattern (Fig. 4) and that 83% of patients with MO cancers were similarly discriminated (Fig. 3). Only 4 (17%) of MO patients were not discriminated from the BPH pattern, and all of these had prostate volumes of less than 50 ml. CONCLUSIONS
An encouraging discriminationwas achieved between patients with BPH and prostate cancer. In this preliminary but retrospective study using Deming regression analysis to compare log-transformed PSA with prostatic volume, all patients with MI disease and 83% of those with MO disease were discriminated from those with BPH. A prospective study to examine in depth the concept of volume adjustment of PSA is being started. The idea could be practically applied as a volume-correctedchart for BPHusedas atemplateand kept forreference inclinics, wards, and scanning rooms, so that when a patient undergoes TRUS volume measurement, a quick comparison of the PSA in relation to the 97.5% confidence limits for that particular volume may give some indication of the significance of the PSA result. Alternatively,the software ofthe scanner could be programmed to indicate this information.
Fig. 2. Correlationbetween prostatevolume and serum prostate specific antigen (PSA) inpatients with benign prostatic hyperplasia. Fig. 3. Prostate volume and serum PSA in patients with non-metastatic prostate cancer in comparison with patients with benign prostatic hyperplasia (BPH). Fig. 4. Prostate volume and serum prostate specific antigen (PSA) in patients with metastatic prostate cancer in comparison with patients with benign prostatic hyperplasia (BPH).
56
Clements et al.
Is correlation between prostate volume and PSA clinically useful? Deming regression analysis of PSA and prostate volume is certainly very promising as a sorting process between patients with BPH and cancer, but its application using normal prostates as reference standards is not yet clear. REFERENCES
1. Oosterom R, Bogdanowicz J, Schrader FH: Evaluation of prostate-specificantigen in untreated prostatic carcinoma. Em Urn1 16253-257,1989. 2. Nadji M, Tabei SZ, Castro A, Chu TM, Murphy GP, Wang MC, Morales AR: Prostatic specificantigen:animmuno-histologicmarkerforprostaticneoplasms.Cancer48:12291232, 1981. 3. Frankel AE, Rouse RV, Wang MC, Chu TM, HerzenbergL A Monoclonal antibodies to a human prostatic antigen. Cancer Res 42:37143718,1982. 4. Stamey TA, Kabalin JN: Prostate specific antigen in the diagnosis and treatment of adenocarcinomaof the prostate I. Untreated patients. J Urol 141:1070-1075,1989. 5. Watanabe H, Igari D, Tanahashi Y, Harada K, Saitoh M: Measurementsof size and weight of prostate by means of transrectal ultrmnotomography. Tohoku J Exp Med 114~227-285,1974.
QUESTIONS
S. Fossil (Oslo, Norway): Professor Peeling, did you say that you had evidence of increased PSA in every M+ disease patient? I have seen M+ patients with a normal PSA. W.B. Peeling: These results were from a group of patients who were studied retrospectively. We now feel encouraged to do a full study and are looking at the idea of a volume-corrected PSA. J.E. Montie (Fort Lauderdale, FL): Inpatients withbenignprostatichyperplasia, there is an unknown amount of cancer present that could contribute to the PSA elevation. How will it be possible to account for that? W.B. Peeling: I don’t think it can be accounted for in a clinical study. Most of the patients on the chart, however, had prostatic volumes of 40 ml or more. In a relatively low volume prostate, which is normal on TRUS, we would be suspicious if PSA was elevated above, for example, 20 ng/ml. We look upon PSA measurement as a means of identifying patients who probably have prostate cancer, and the area of particular interest is 4-10 ng/ml. S.Fossil: Do you thinkthat PSA really discriminates patients with pathological M+ disease? Many of the patients with advanced MO disease will have lymph node involvement, thus confusing the interpretation of the results in this group. W.B. Peeling: It is hoped that diagnostic processes using TRUS and MRI will improve the sensitivity of imaging and also that the PSA assay will improve with
Correlations of Prostate Volume and PSA
57
time. At the moment, we are working with a normal range of 0-4ng/ml (Hybritech). If the sensitivity of the process was increased so that the normal range was 0-100, the technique may become more accurate. F.K.Mostofi (Washington, D.C.):Not every prostate cancer produces PSA. There are cancers of the prostate, particularly poorly differentiated cancers, which irrespective of their volume do not produce PSA.
