- Email: [email protected]
Measurement of the proportion of free to total prostate-specific antigen improves diagnostic performance of prostate-specific antigen in the diagnostic gray zone of total prostate-specific antigen
MEASUREMENT OF THE PROPORTION OF FREE TO TOTAL PROSTATE-SPECIFIC ANTIGEN IMPROVES DIAGNOSTIC PERFORMANCE OF PROSTATE-SPECIFIC ANTIGEN IN THE DIAGNOSTIC GRAY ZONE OF TOTAL PROSTATE-SPECIFIC ANTIGEN ALBERT A. LUDERER, PH.D., YA-TING CHEN, PH.D., THOMAS E SORIANO, B.S., WILLIAM J. KRAMP, PH.D., GRANT CARLSON, B.S., CAROL CUNY, B.S., THOMAS SHARP, B.S., WILLIAM SMITH, B.S., JASON PETTEWAY, M.S., MICHAEL K. BRAWER, M.D., AND ROBERT THIEL, PH.D.
ABSTRACT-Objectives. This study examined the clinical significance of non-complexed (free) prostate-specific antigen (PSA) in the differential diagnosis of prostate cancer with an emphasis on patients with total PSA values between 4.0 and 10.0 ng/mL (the diagnostic gray zone]. Methods. Serum samples were obtained from three specimen banks. Patient samples consisted of 55 untreated histologically confirmed primary cancer, 62 men with untreated benign prostatic disease histologically confirmed by 6 negative sextant biopsies, and 64 asymptomatic healthy male controls with normal digital rectal examinations and PSA values less than 4.0 ng/mL. All patients were between the ages of 50 and 75 years. Total PSA levels were determined using the PA immunoassay performed on the TOSOH AIA- 1200 automated immunoassay instrument. Free PSA levels were determined using a monoclonal-polyclonal antibody sandwich radioimmunoassay. The proportion of free to total PSA was calculated by dividing the patient’s free PSA value by the total PSA value. Results. When all subjects were included, both total PSA and the proportion of free to total PSA significantly differentiated between patients with prostate cancer and patients with benign histologic conditions (P c 0.0001). However, in men with total PSA values between 4.0 and 10.0 ng/mL, the proportion of free to total PSA significantly differentiated between patients with benign and malignant histologic conditions (P = 0.0004), whereas the total PSA did not (P = 0.13). Among this subgroup of patients, the analysis of sensitivity and specificity showed that the proportion of free to total PSA had a clearly higher specificity compared with that of the total PSA at the same level of sensitivity. Conclusions. Measurement of the free PSA level in a patient’s serum and calculation of the proportion of free to total PSA enhances the ability to distinguish benign histologic conditions from cancer while retaining high sensitivity for detecting cancer in men who present with total PSA levels between 4.0 and 10.0 ng/mL. A large-scale population-based study is currently in progress to confirm this preliminary finding. UROLCGY’ 46: 187-194, 1995.
rostate-specific antigen (PSA), a single-chain 33-kd glycoprotein serine protease,’ is widely used as a clinical marker of prostate cancer. Although PSA has provided clinicians with an additional useful tool for detecting prostate cancer,
P
From the Department of Research and Development, Dianon Systems, Inc., Stratford, Connecticut; Department of Urology, University of Washington Medical Center and Seattle Veterans Affairs Administration Medical Center, Seattle, Washington; and Department of Public Health, Southern Connecticut State University, New Haven, Connecticut Albert A. Luderer, Ph.D., Department of Research and Development, Dianon Systems, Inc., Stratford, CT 06497 Submitted (Rapid Communication): March 20, 199.5, accepted (with revisions): April 7, 1995 UROLOGY@ 46 (2), 1995
the sensitivity and specificity of PSA are not sufficient to make it an ideal tumor marker for screening and early detection. Generally speaking, PSA is organ-specific but not cancer-specific, and above-normal serum PSA levels occur in a high proportion of patients with benign prostatic hyperplasia (BPH) .’ Recently, studies have found that there are different molecular forms of PSA (free and complexed) circulating in the bloodstream.3-5 Complexed PSA is predominantly bound to a,antichymotrypsin (PSA-ACT). This form is measured in all commercial assays for PSA. PSA is also complexed to a,-macroglobulin; however, this form is not detected in current commercial assays 187
for PSA. Noncomplexed (free) PSA is present in much lower concentrations than complexed PSA and is measured in varying degrees by all commercial PSA immunoassays. Depending on the immunoassay used to measure serum PSA, various amounts of free and complexed PSA are recognized and contribute to the total measurable PSA.6 Little is known about the clinical significance of free or complexed PSA. A number of studies reported the potential of increasing the sensitivity and specificity for diagnosing prostate cancer using free or complexed PSA or their proportions to total PSA.5,7-g Christensson et ~1.~ found that the percentage of free PSA was significantly lower in patients with untreated prostate cancer than in those with benign disease. Utilizing a total PSA cutoff of 5 ng/mL, Christensson et al. reported a clinical sensitivity and specificity of 90% and 55%, respectively. However, when the total PSA and the proportion of free to total PSA were combined at 10 ng/mL and 0.18, respectively, specificity increased to 73% without affecting sensitivity. This data set included serum from cancer patients with PSA levels as great as thousands of nanograms per milliliter and did not address the differential diagnostic dilemma that occurs in the 4 to 10 ng/mL range. Other studies5,8*g showed that levels of complexed PSA were higher in patients with prostate cancer than in patients with benign disease or normal controls. None of the previously cited studies evaluated the free or complexed forms of PSA or their proportions within the so-called diagnostic gray zone (total PSA 4.0 to 10.0 ng/mL). The current study evaluated the clinical significance of the free form of PSA and its proportion to total PSA in the differential diagnosis of prostate cancer, with an emphasis on the total PSA range of 4.0 to 10.0 ng/mL. MATERIAL PATIENTACCRUALS
AND
METHODS
Serum samples were obtained from three specimen banks that collect samples from clinical practices across the United States. All patients who satisfied the inclusion criteria for the study were identified from each of the specimen banks. Patient inclusion criteria for this study were as follows: (1) serum samples were obtained within 6 weeks prior to any physical manipulation of the prostate (digital rectal examination [DRE] or biopsy); (2) patients had no history of prostate diseaserelated therapeutic intervention; (3) patient disease status was histologically confirmed by six-sextant biopsy; and (4) patients had no history of cancer. Asymptomatic healthy normal patients who had normal PSA values and DRE were identified as the control group. Because of the potential confounding effect of age on the performance of total PSA,‘O only patients between 50 and 75 years old were included in the current analysis to ensure a more homogeneous patient population. There were 55 patients with prostate cancer, 62 patients with benign conditions, and 64 asymptomatic healthy male control subjects who met the selection criteria and were included in the current analysis. All blood samples were stored at less than -20°C until assayed. Gleason score and clinical stage (TNM 188
J 0.8 E 2 c 0.6
;:
a 0.4 0.2
0 1. Superose-6 high-performance chromatography fraction analysis of molecular forms of PSA. FIGURE
stage grouping) for prostate cancer patients were obtained from the pathology reports and medical charts, respectively.
The level of total PSA in serum samples was measured using the PA immunoassay (a Food and Drug Administration-approved assay) performed on the TOSOH-AIA 1200 automated immunoassay analyzer following the recommended protocol of the manufacturer. This immunoassay is a dual monoclonal sandwich assay that recognizes both the free form of PSA (approximately 30 kd) and the complexed form of PSA (approximately 90 kd) (Fig. 1). The TOSOH PSA analytic sensitivity was 0.1 ng/mL, which represents 2 standard deviations over the mean optical density of the assay zero calibrator. Inter- and intra-assay coefficients of variation were less than 10%. The assay was linear across its measuring range (0.0 to 100.0 ng/mL). The amount of free PSA in serum samples was measured using an investigational double antibody immunoradiometric assay available through Dianon Systems, Stratford, Connecticut.” This assay recognizes only the free, uncomplexed form of PSA. Assay components consisted of a monoclonal anti-free PSA antibody (Immunocorp Sciences, Montreal, Canada) coated onto polystyrene 12 X 75 mm tubes; affinity purified, rabbit polyclonal anti-PSA labeled with lz51 and free PSA standards created from human seminal plasma.‘* The mass of free PSA was confirmed by both TOSOH and Yang PSA methods. Standards were prepared by the addition of a known mass of human seminal plasma-free PSA to a protein matrix buffer consisting of 0.01 M phosphate-buffered saline, 0.25% bovine serum albumin, and 0.01% sodium azide (NaN,). Assays were performed at 23°C. One hundred microliter of neat serum and 100 pL of assay buffer were added to the coated tubes and incubated for 2 hours with constant low-speed shaking. Tubes were washed two times with 4 mL of wash buffer (0.01 M phosphate buffered saline, 0.25% bovine serum albumin, 0.01% Tween 20, and 0.01% NaN,) followed by the addition of 200 pL of lz51 anti-PSA. After 2 hours of incubation with low-speed shaking, tubes were twice washed with 4 mL of wash buffer, decanted, and the total bound radioactivity determined in a gamma counter. The amount of free PSA specifically captured by the solidphase monoclonal anti-free PSA was estimated by extrapolation of the amount of 1251 tracer bound against a standard curve. A free PSA standard curve representing two separate lots of iodinated tracer is illustrated in Figure 2. Specificity was demonstrated by reactivity profile analysis of fractionated patient specimens. Male serum samples with known total PSA values were chromatographed by high-performance liquid chromatography on a Superose-6 I.Ea.oGY@
46 (2), 1995
0.1. F-PSA
Dilution
(ng/mL)
Typical free prostate-specific antigen (FPSA) calibration curve. Dynamic range of 0.0 to 50.0 ng/mL. Plotted are the mean f 2 SD of each standard from four separate runs and from two separate tracer lots (lot l[..V...], lot 2 [-*-I). B/T% is the percentage of bound counts per minute (CPM) over the total CPM added. FIGURE 2.
(Pharmacia, Piscataway, NJ) support under nondenaturing conditions. Resultant fractions were analyzed for total and free PSA. The total PSA reactivity profile of these fractionated serum specimens clearly showed two discrete peaks of activity: the first at approximately 90 kd corresponding to the PSA-ACT complex and the second at approximately 30 kd representing the free PSA (Fig. 1). The free PSA assay’s selectivity for free PSA was demonstrated repeatedly in both native serum specimens and also with artificially prepared samples of total PSA containing different proportions of free PSA to PSA-ACT (data not shown).
TABLE
Sample Nl*
N2 N3 N4 N5 Pl P2 P3 P4 P5 Mean percent recovery*
1 3.92* 0.16 0.25 0.37 0.10 6.45 0.43 0.69 2.16 1.27
10
1
I.
Freeze/thaw
Factor
FIGURE 3. Serial dilution curves of three cancer patient serum samples. Five dilutions for each sample. The mean value derived from the average of four replicates. Sample 1 b, mean observed values; - -, expected dilution line) y = 1. lx - 0.19, r = 0.99. Sample 2 (m,mean observed values; ...... expected dilution line) y = 1.0 lx + 0.59, r = 0.99. Sample 3 (A, mean observed values;-..-, expected dilution line) y = 1.18x + 0.32, r = 0.99. The free PSA assay demonstrated an analytical sensitivity range of 0.02 to 0.04 ng/mL, which was determined by comparing the mean of the lowest standard (0.2 ng/mL) and the 99% confidence interval of the assay’s zero calibrator over 10 replicates performed on 3 different days. The free PSA assay possesses an acceptably low total coefficient of variation, 10% or less, for combined inter- and intra-run variability across the measuring range of the assay. The assay was linear across its measuring range (0.0 - 50.0 ng/mL) (Fig. 3). Free PSA levels in specimens were not affected by up to 10 repeated minus 70” to 23°C freeze/thaw cycles (Table I) and
stability
Cycle Number 2 6 4.39 4.56 0.22 0.22 0.25 0.29 0.38 0.41 0.11 0.16 6.14 6.31 0.43 0.45 0.71 0.70 2.15 2.07 1.30 1.24 106 114
of the free PSA antigen 10 4.53 0.20 0.32 0.41 0.13 7.07 0.45 0.74 2.08 1.29 113
Overall Mean 4.35 0.20 0.28 0.39 0.13 6.49 0.44 0.7 1 2.12 1.28
cv (%) 6.8 14.1 12.3 5.3 21.2 6.2 2.6 3.0 2.2 2.1
KEY: CV = coefficient of variation. ‘Sample type: N = normal sewn samples; P = prostate cancer patient serum samples. ‘Values represent the average of4 replicate determinations offree PSA in ng/mL. *Determined by the comparison ofeach additional cycle tocycle I.
UROLOGYa 46 (2), 1995
189
TABLE II. Sample Nl* N2 N3 N4 N5 Pl P2 P3 P4 P5 Mean percent recovery*
Day-to-day
stability
of tee free PSA antigen
Cycle Number 3 2 3.88 4.09 0.18 0.22 0.25 0.24 0.40 0.44 0.08 0.12 9.53 9.58 2.25 2.21 0.46 0.46 1.91 1.88 0.92 0.89 104 105
1 3.92+ 0.16 0.25 0.37 0.10 9.65 2.30 0.39 1.95 0.90
6 4.32 0.22 0.3 1 0.30 0.16 9.22 2.15 0.54 1.97 0.92 114
Overall Mean 4.05 0.20 0.26 0.38 0.12 9.50 2.23 0.46 1.93 0.91
cv % 4.9 15.4 12.2 15.7 29.7 2.0 2.8 13.3 2.1 1.7
KEY: CV = coefficient of variation. *Sample type: N = normal serum samples; P = prostate cancer patient senm samples. ‘Values represent the average of4 replicate determinations offree PSA in ng/mL. *Determined by the comparison ofeach additional cycle to cycle I.
TABLE III.
Sample No. 1 2 3 4
Evaluation
Control FPSA (n&W 0.43 0.96 1.41 8.37
of the effect of intetiering Hemoglobin (500 mgldl) FPSA (w/mL) 0.39 1.00 1.40 8.20
Recovery* WI 90.7 104.2 99.3 98.0
substances
in the free PSA immunoassay
Bilirubin (20 mg/dL) FPSA Recovery WmLl (%I 0.38 88.4 1.02 106.3 1.37 97.2 100.0 8.37
(1000 FPSA (n&W 0.38 1.03 1.43 8.25
Lipid mg/dL] Recovery (%I 88.4 107.3 101.4 98.6
KEY: FPSA =free PSA. *Recovery determined with assay buffer as the control.
exhibited good room temperature stability over six repetitive temperature cycles between 23°C for 30 minutes followed by 24 hours at 4’C (Table II). Interfering substances were evaluated by the addition of known amount of interferents to serum pools containing known amount of free PSA. Controls were corrected for the volume effect. Table III demonstrates that none of the potential interferent substances (hemoglobin 500 mg/dL, bilirubin 20 mg/dL, lipid 1000 mg/dL) interfered with the accurate determination of free PSA.
DATA COLLECTION AND DATA ANALYSIS Total and free PSA tests were performed at Dianon Systems, with operators blinded to the clinical specimen type (cancer, benign, control). The proportion of free PSA in a patient’s sample was calculated by dividing the patient’s free PSA value by the total PSA value. This proportion, expressed as a percent, will be referred to as free/total PSA throughout the remainder of this article. A structured data abstract form was designed to obtain clinical information on the patient’s age, history of prostate disease, history of prostate-related treatment, history of cancer, and needle biopsy results. Data were abstracted from the computer database in each of the three specimen banks. A nonparametric procedure (Wilcoxon) was used to test the statistical significance of differences in the median values of total PSA and free/total PSA between BPH and prostate cancer patients. Sensitivity and specificity were calculated. Receiver-operating characteristic (ROC) curves were generated by plotting sensitivity versus specificity to evaluate the assay performance for total PSA and for free/total PSA. 190
RESULTS Frequency distributions of patient age at disease diagnosis, total PSA levels, and the free/total PSA levels are presented in Table IV. The majority of the patients, 68% with cancer, 72% with BPH, and 52% asymptomatic normal, were between the ages of 55 and 64 years. The mean age was 64 years for cancer patients, 62 years for men with BPH, and 62 years for asymptomatic normal control subjects. Distribution of total PSA levels was notably different among the three patient groups. All asymptomatic normal controls had total PSA levels below 4.0 ng/mL; 23 (37%) patients with BPH were below 4.0 ng/mL, and only 4 (7%) cancer patients had a total PSA below 4.0 ng/mL. More cancer patients (73%) had low levels of free/total PSA (less than 15%) than patients with BPH (29%) or normal control subjects (24%). Among cancer patients, the distribution of the Gleason score was as follows: 58% (32 of 55) Gleason 5 or 6, 24% (13 of 55) Gleason 7 or 8, 11% (6 of 55) Gleason 9 or 10, and 7% (4 of 55) Gleason score not available. Information on clinical stage, based on TNM stage grouping, was available for 71% (39 of 55) of the cancer patients. URoLmYm 46 (2), 1995
Frequency distributions of patient age at disease diagnosis, total PSA * and proportion of free to total PSA by patient group
TABLE
IV.
Patient Cancer No. (%) 55 (1001
Total samples Age (yrl 50-54 55-59 60-64 65-69 70-75 Total PSA (ng/mL] <4 4-10 >lO Free/total PSA (Oh) 510 10.1-15 15.1-20 20.1-25 25.1-30 >30
1 7 30 9 8
(2) (131 (55)
(161 (151
Group
Benign No. (%) 62 (100)
Normal Control No. (%) 64 (1 OO]
4 17 28 5 8
11 10 23 13 7
(61 (27) (45) (8) (131
(17) (16) (36) (201 (1 11
4 (71 25 (45) 26 (47)
23 (37) 32 (52) 7 (1 1)
64 (100) 0 (01 0 (0)
19 (35) 21 (38) 8 (151
7 11 12 11 7 14
3 12 10 16 5 18
6 (1 11 1 (2) 0 (01
(111 (18) (19) (18) (1 11 (23)
(51 (19) (16) (25) (8) (28)
* TOSOH immunoassay
Among these patients, 11 were Stage I, 21 were Stage II, 3 were Stage III, and 4 were Stage IV Descriptive statistics for total PSA and the free/total PSA are summarized in Tables V and VI. When all subjects were included in the analysis (Table V), the median value for total PSA was highest among cancer patients (9.60 ng/mL), followed by patients with BPH (4.89 ng/mL), and was lowest among normal control subjects (1.07 ng/mL). A nonparametric Wilcoxon test indicated that the
differences in the medians of total PSA between cancer and BPH patients were statistically significant (P
Descriptive statistics for total PSA * and proportion free to total PSA for all subjects by patient group
TABLE V.
Patient Normal Control (n = 64) Total PSA (ng/mL) Mean Median Standard deviation 95% Cl for mean Free/Total PSA (%) Mean Median Standard deviation 95% Cl for mean KEY: 95% CI = 95 percent * TOSOH immunoassay. ‘Wilcoxon test comparing
confidence median
UROLOGYa 46 (2), 199.5
1.24 1.07 0.81 1.03- 1.44 26 23 14 23-30 interval values
Benign (n = 62)
of
Group Cancer (n = 55)
5.52 4.89 4.12 4.47-6.75
13.64 9.60 15.18 9.54-l 7.75
23 21 13 20-27
13 13 6 11-14
P Value’
< 0.000 1
< 0.000 1
of the mean.
between
benign
and cancer patients.
191
TABLE VI.
Descriptive
statistics for total PSA* and proportion of free to total PSA for subjects with total PSA 4.0 to 10.0 ng/mL by patient group Patient Croup Cancer (n = 25)
Benign (n = 32) Total PSA (ng/mL) Mean Median Standard deviation 95% Cl for mean Free/Total PSA (%) Mean Median Standard deviation 95% Cl for mean
6.18 6.05 1.48 5.65-6.72
P Value+
6.92 6.90 1.75 6.20-7.65
0.13
13 14 6 11-16
22 21 9 18-25
0.0004
KEY: 95% CI = 95 percentconfidence intervalfor the mean. * TOSOH immunoassay. tWilcoxon test comparing median values between benign and cancer patients.
100
80
0-l 0
FIGURE analysis free and primary disease.
/ 20
40 l-specificity
60
80
100
4. Receiver-operating characteristic curve comparing total PSA and the proportion of total (F/T) PSA based on 55 patients with prostate cancer and 62 patients with benign All PSA values are included.
clinical database, analysis revealed an improved performance for free/total PSA compared to total PSA (Fig. 5; Table VII). Free/total PSA had a clearly higher specificity than total PSA for almost all sensitivity levels in the total PSA range of 4.0 to 10.0 ng/mL (Fig. 5; Table VII). This observation was more profound when the value of free/total PSA was higher. At free/total PSA values 25% or less, all cancers were detected, and 31% of the patients with BPH were correctly identified. This selectivity is in contrast to total PSA 4.0 ng/mL or higher, which correctly identified all cancer patients but could not identify any patients with BPH (Table VII). COMMENT
The objective uate the clinical 192
of the current study was to evalutility of free PSA to differentiate
0, : I ;-0
20
40 I -epecif$
1 : : IA 80
100
FIGURE 5. Receiver-operating characteristic (ROC) curve analysis comparing total PSA and the proportion of free and total (F/r) PSA based on 25 patients with primary prostate cancer and 32 with benign disease with total PSA values between 4 and 10 ng/mL.
between benign and malignant prostate disease in the diagnostic gray zone of total PSA values 4.0 to 10.0 ng/mL. Similar to Christensson et al.,7 we observed an overall significantly smaller proportion of free PSA in prostate cancer patients than in patients with BPH. More importantly, when data were limited to a range of 4.0 to 10.0 ng/mL total PSA, differences in the proportions of free to total PSA between cancer and benign patients remained statistically significant, whereas differences in total PSA values were not detectable (Table VI). The PSA test has become one of the most costeffective tools for detecting early prostate cancer. 2~13~14In general, a diagnosis of prostate cancer is uncommon at serum levels of total PSA at or below 4 ng/mL and is common at levels above 10 ng/mL. The diagnostic gray zone is between 4.0 and 10.0 ng/mL, where the differential diagnosis uROLoGYrn 46 (21, 1995
TABLE VII.
Sensitivity
and specificity for different total PSA and proportion values for subjects with total PSA 4.0 to 10.0 ng/mL
Sensitivity
[%)
(NoJtotal)
Specificity
of free to total PSA (%)
(NoJtotal)
Total PSA
(w/m Ll 24 25 26 27 >8 29 Free/Total PSA (%) 530 525 520 515 514 112 510 57
100.0 92.0 68.0 48.0 36.0 16.0
(25/25) (23/25) (17125) (12125) (9/25) (4/25)
0.0 28.1 50.0 65.6 90.6 93.8
(O/32) (9/32) (16132) (21/32) (29/32] (30132)
100.0 100.0 88.0 72.0 68.0 36.0 24.0 20.0
(25/25) (25/25) (22/25) (18125) (17125) (9/25) (6125) 15/251
15.6 31.3 50.0 75.0 81.3 81.3 87.5 100.0
(5/32) (10/32) (16/32) (24/32) (26/32) (26/32) (28/32) [32/32)
of prostate cancer is most difficult2J4 Our data showed similar results to prior studies in that there were significant differences in total PSA distribution between patient groups over all values of total PSA, but not in the range between 4.0 and 10.0 ng/mL (Tables V VI). Alternative methods for increasing the sensitivity and specificity of total PSA within this diagnostic gray zone have been proposed such as PSA density,15 PSA velocity,16 and age-specific reference range. l7 PSA density and PSA velocity have contributed significantly to disease understanding but have suffered practical limitation due to the difficulty of accurately measuring prostate volume for PSA density calculations and the inherent biologic and laboratory variability associated with serial PSA determinations required for PSA velocity calculations1*~r9 Agespecific reference ranges should increase the sensitivity of PSA in younger men and the specificity of PSA in older men.20 Our study results suggest that the proportion of free to total PSA has intrinsic value as an aid in the differential diagnosis of benign disease from cancer of the prostate in the range of 4.0 to 10.0 ng/mL total PSA (Table VI; Fig. 5). Based on our analysis, free/total PSA as a single test did not differentiate between benign and prostate cancer patients significantly better than the total PSA alone over the entire PSA range (Fig. 4). If free/total were to be used as a single test, the optimal sensitivity and specificity would be 87% and 52%, respectively, at a cutoff of 20% free/total PSA based on the current data set. Compared with a total PSA cutoff of 4 ng/mL, free/total PSA improved the specificity from 37% to 52%; however, sensitivity worsened from 93% to 87%. As a sinUROLOGY@ 46 121, 1995
y/ /-l
4
Eknign / I 2l
Ii
40% Percent
60% Free /Total
80%
100%
PSA
FIGURE 6. Theoretical distributions for the proportion of free to total PSA in patients with prostate cancer and benign histologic results.
gle test, the sensitivity and specificity of a free/total PSA cutoff of 20% was essentially identical to a total PSA cutoff of 5 ng/mL. Theoretical distributions for free/total PSA values in subjects with total PSA values in the range of 4.0 to 10.0 ng/mL were constructed from our database (Fig. 6). These distributions were generated using curve fitting software.21 (To determine these distributions, we first fitted curves to our empirical cumulative data in each patient group. This process generated many different equations that fitted the data. Among all equations, a functional form known as the log-normal cumulative function was chosen because the equation had a very good fit [r2 >0.951 in both patient groups, and the functional equation is biologically plausible. That is, the shape of each distribution is similar to other biologically controlled measures such as blood pressure and 193
hematocrit. Once we identified the functional form of the distribution, the probability density function for each patient group can be easily derived. It is the probability density function that is plotted in Figure 4. The units on the vertical axis are arbitrary. They are derived from the fact that the total area under each curve must equal unity.) At a cutoff of 20% free/total PSA, the theoretical curves show a sensitivity of 80% and a specificity of 49%. These values are within the 95% confidence interval of our data estimate of a sensitivity of 88% and a specificity of 50% at the same free/total PSA cutoff. The figure clearly shows the effective separation that can be attributed to the proportion of free/total PSA in the diagnostic gray zone of total PSA. One of the most interesting observations in our total PSA 4 to 10 ng/mL subdata set was the absence of prostate cancer in patients with free/total PSA more than 25%. Thirty percent of all patients with benign conditions whose original total PSA values lay between 4.0 and 10.0 ng/mL fell into this category (Table VII). Age is an important factor affecting prostate gland volume and the prostate’s ability to produce and secrete PSA.1° Both PSA level and the volume of the prostate gland increase with increasing age. lo In the current study, we were able to reduce the potential confounding effect of age by restricting patient’s age to the range of 50 to 75 years. The frequency distribution of age was similar among the three patient groups, although the mean age was slightly greater in prostate cancer patients (64 years) than in patients with BPH, or healthy normal control subjects (62 years) (Table IV). Misclassification of patient status, particularly patients with benign prostate disease, was unlikely in the current study because sextant biopsies were performed for patient disease confirmation. REFERENCES 1. Wang MC, Valenzuela LA, Murphy GP, and Chu TM: Purification of a human prostate-specific antigen. Invest Urol 17: 159-163, 1979. 2. Stamey TA, Yang N, Hay AR, McNeil JE, Freiha FS, and Redwine E: Prostate-specific antigen as a marker for adenocarcinoma of the prostate. N Engl J Med 317: 909-916, 1987. 3. Christensson A, Laurel1 CB, and Lilja H: Enzymatic activity of the prostate-specific antigen and its reactions with extra-cellular serine proteinase inhibitors. Eur J Biochem 194: 755-763, 1990. 4. Lilja H, Christensson A, Dahlen U, Matikainen M, Nilsson 0, Petterson K, and Lovgren T: Prostate-specific antigen in serum occurs predominantly in complex with a,antichymotrypsin. Clin Chem 37: 1618-1625, 1991. 5. Stenman UH, Leinonen J, Alfthan H, Rannikko S, Tuhkanen K, and Alfthan 0: A complex between prostate-
specific antigen and a,-antichymotrypsin is the major form of prostate-specific antigen in serum of patients with prostatic cancer: assay of the complex improves clinical sensitivity for cancer. Cancer Res 51: 222-226, 1991. 6. Partin AW, and Oesterling JE: The clinical usefulness of prostate specific antigen: update 1994. J Urol 152: 1358-1368, 1994. 7. Christensson A, Bjork T, Nilsson 0, Dahlen U, Matikainen MT, Cockett AT, Abrahamsson PA, and Lilja H: Serum prostate-specific antigen complexed to ai-antichymotrypsin as an indicator of prostate cancer. J Urol 150: 100-105, 1993. 8. Wood WG, van der Sloot E, and Bohle A: The establishment and evaluation of luminescent-labelled immunometric assays for prostate-specific antigen-a,-antichymotrypsin complexes in serum. Eur J Chem Clin Biochem 29: 787-794, 1991. 9. Leinonen J, Lovgren T, Vornanen T, and Stenman UH: Double-label time-resolved immunofluorometric assay of prostate-specific antigen and of its complex with c~iantichymotrypsin. Clin Chem 39: 2098-2103, 1993. 10. Oesterling JE, Cooner WH, Jacobsen SJ, Guess HA, and Lieber MM: Influence of patient age on the serum PSA concentration. An important clinical observation. Urol Clin North Am 20: 671-680, 1993. 11. Cuny C, Pham L, Kramp W, Sharp T, and Soriano TF: Evaluation of a two-site immunoradiometric assay for the measurement of non-complexed (free) PSA. Clin Chem 1995. (submitted) 12. Rusciano D, Berardi A, Ceccarini C, and Terrana B: Concomitant purification of prostatic carcinoma tumor markers from human seminal fluid under nondenaturing conditions Clin Chem 34: 2528-2532, 1988. 13. Brawer MK, Chetner MP, Beatie J, Buchner DM, Vessella RL, and Lange PH: Screening for prostatic carcinoma with prostate specific antigen. J Urol 147: 841-845, 1992. 14. Catalona WJ, Smith DS, Ratliff TL, Dodds KM, Coplen DE, Yuan JJ, Petros JA, and Andriole GL: Measurement of prostate-specific antigen in serum as a screening test for prostate cancer. N Engl J Med 324: 1156-1161, 1991. 15. Benson MC, and Olsson CA: Prostate specific antigen and prostate specific antigen density. Roles in patient evaluation and management. Cancer 74: 1667-1673, 1994. 16. Carter HB, Pearson JD, Metter EJ, Brant LJ, Chan DW, Andres R, Fozard JL, and Walsh PC: Longitudinal evaluation of prostate specific antigen levels in men with and without prostate disease. JAMA 267: 2215-2220, 1992. 17. Oesterling JE, Jacobsen SJ, Chute CG, Guess HA, Girman CJ, Panser LA, and Lieber MM: Serum prostate-specific antigen in a community-based population of healthy men. Establishment of age-specific reference ranges. JAMA 270: 860-864, 1993. 18. Mettlin C, Littrup PJ, Kane RA, Murphy GP, Lee F, Chesley A, Badalament R, and Mostofi FK: Relative sensitivity and specificity of serum prostate specific antigen (PSA) level compared with age-referenced PSA, PSA density, and PSA change. Cancer 74: 1615-1620, 1994. 19. Stamey TA: Second Stanford Conference on International Standardization of Prostate-Specific Antigen Immunoassays: September 1 and 2, 1994. Urology 45: 173-184, 1995. 20. Catalona WJ, Hudson MA, Scardino PT, Richie JP, Ahmann FR, Flanigan RC, deKernion JB, Ratliff TL, Kavoussi LR, Dalkin BL, et al: Selection of optimal prostate specific antigen cutoffs for early detection of prostate cancer: receiver operating characteristic curves. J Urol 152: 2037-2042, 1994. 21. Jandel Scientific: TableCurve 20 for Windows. AISN Software Inc., San Raphael, California, 1994.
UROLOGY’
46 (21, I995
ABSTRACT-Objectives. This study examined the clinical significance of non-complexed (free) prostate-specific antigen (PSA) in the differential diagnosis of prostate cancer with an emphasis on patients with total PSA values between 4.0 and 10.0 ng/mL (the diagnostic gray zone]. Methods. Serum samples were obtained from three specimen banks. Patient samples consisted of 55 untreated histologically confirmed primary cancer, 62 men with untreated benign prostatic disease histologically confirmed by 6 negative sextant biopsies, and 64 asymptomatic healthy male controls with normal digital rectal examinations and PSA values less than 4.0 ng/mL. All patients were between the ages of 50 and 75 years. Total PSA levels were determined using the PA immunoassay performed on the TOSOH AIA- 1200 automated immunoassay instrument. Free PSA levels were determined using a monoclonal-polyclonal antibody sandwich radioimmunoassay. The proportion of free to total PSA was calculated by dividing the patient’s free PSA value by the total PSA value. Results. When all subjects were included, both total PSA and the proportion of free to total PSA significantly differentiated between patients with prostate cancer and patients with benign histologic conditions (P c 0.0001). However, in men with total PSA values between 4.0 and 10.0 ng/mL, the proportion of free to total PSA significantly differentiated between patients with benign and malignant histologic conditions (P = 0.0004), whereas the total PSA did not (P = 0.13). Among this subgroup of patients, the analysis of sensitivity and specificity showed that the proportion of free to total PSA had a clearly higher specificity compared with that of the total PSA at the same level of sensitivity. Conclusions. Measurement of the free PSA level in a patient’s serum and calculation of the proportion of free to total PSA enhances the ability to distinguish benign histologic conditions from cancer while retaining high sensitivity for detecting cancer in men who present with total PSA levels between 4.0 and 10.0 ng/mL. A large-scale population-based study is currently in progress to confirm this preliminary finding. UROLCGY’ 46: 187-194, 1995.
rostate-specific antigen (PSA), a single-chain 33-kd glycoprotein serine protease,’ is widely used as a clinical marker of prostate cancer. Although PSA has provided clinicians with an additional useful tool for detecting prostate cancer,
P
From the Department of Research and Development, Dianon Systems, Inc., Stratford, Connecticut; Department of Urology, University of Washington Medical Center and Seattle Veterans Affairs Administration Medical Center, Seattle, Washington; and Department of Public Health, Southern Connecticut State University, New Haven, Connecticut Albert A. Luderer, Ph.D., Department of Research and Development, Dianon Systems, Inc., Stratford, CT 06497 Submitted (Rapid Communication): March 20, 199.5, accepted (with revisions): April 7, 1995 UROLOGY@ 46 (2), 1995
the sensitivity and specificity of PSA are not sufficient to make it an ideal tumor marker for screening and early detection. Generally speaking, PSA is organ-specific but not cancer-specific, and above-normal serum PSA levels occur in a high proportion of patients with benign prostatic hyperplasia (BPH) .’ Recently, studies have found that there are different molecular forms of PSA (free and complexed) circulating in the bloodstream.3-5 Complexed PSA is predominantly bound to a,antichymotrypsin (PSA-ACT). This form is measured in all commercial assays for PSA. PSA is also complexed to a,-macroglobulin; however, this form is not detected in current commercial assays 187
for PSA. Noncomplexed (free) PSA is present in much lower concentrations than complexed PSA and is measured in varying degrees by all commercial PSA immunoassays. Depending on the immunoassay used to measure serum PSA, various amounts of free and complexed PSA are recognized and contribute to the total measurable PSA.6 Little is known about the clinical significance of free or complexed PSA. A number of studies reported the potential of increasing the sensitivity and specificity for diagnosing prostate cancer using free or complexed PSA or their proportions to total PSA.5,7-g Christensson et ~1.~ found that the percentage of free PSA was significantly lower in patients with untreated prostate cancer than in those with benign disease. Utilizing a total PSA cutoff of 5 ng/mL, Christensson et al. reported a clinical sensitivity and specificity of 90% and 55%, respectively. However, when the total PSA and the proportion of free to total PSA were combined at 10 ng/mL and 0.18, respectively, specificity increased to 73% without affecting sensitivity. This data set included serum from cancer patients with PSA levels as great as thousands of nanograms per milliliter and did not address the differential diagnostic dilemma that occurs in the 4 to 10 ng/mL range. Other studies5,8*g showed that levels of complexed PSA were higher in patients with prostate cancer than in patients with benign disease or normal controls. None of the previously cited studies evaluated the free or complexed forms of PSA or their proportions within the so-called diagnostic gray zone (total PSA 4.0 to 10.0 ng/mL). The current study evaluated the clinical significance of the free form of PSA and its proportion to total PSA in the differential diagnosis of prostate cancer, with an emphasis on the total PSA range of 4.0 to 10.0 ng/mL. MATERIAL PATIENTACCRUALS
AND
METHODS
Serum samples were obtained from three specimen banks that collect samples from clinical practices across the United States. All patients who satisfied the inclusion criteria for the study were identified from each of the specimen banks. Patient inclusion criteria for this study were as follows: (1) serum samples were obtained within 6 weeks prior to any physical manipulation of the prostate (digital rectal examination [DRE] or biopsy); (2) patients had no history of prostate diseaserelated therapeutic intervention; (3) patient disease status was histologically confirmed by six-sextant biopsy; and (4) patients had no history of cancer. Asymptomatic healthy normal patients who had normal PSA values and DRE were identified as the control group. Because of the potential confounding effect of age on the performance of total PSA,‘O only patients between 50 and 75 years old were included in the current analysis to ensure a more homogeneous patient population. There were 55 patients with prostate cancer, 62 patients with benign conditions, and 64 asymptomatic healthy male control subjects who met the selection criteria and were included in the current analysis. All blood samples were stored at less than -20°C until assayed. Gleason score and clinical stage (TNM 188
J 0.8 E 2 c 0.6
;:
a 0.4 0.2
0 1. Superose-6 high-performance chromatography fraction analysis of molecular forms of PSA. FIGURE
stage grouping) for prostate cancer patients were obtained from the pathology reports and medical charts, respectively.
The level of total PSA in serum samples was measured using the PA immunoassay (a Food and Drug Administration-approved assay) performed on the TOSOH-AIA 1200 automated immunoassay analyzer following the recommended protocol of the manufacturer. This immunoassay is a dual monoclonal sandwich assay that recognizes both the free form of PSA (approximately 30 kd) and the complexed form of PSA (approximately 90 kd) (Fig. 1). The TOSOH PSA analytic sensitivity was 0.1 ng/mL, which represents 2 standard deviations over the mean optical density of the assay zero calibrator. Inter- and intra-assay coefficients of variation were less than 10%. The assay was linear across its measuring range (0.0 to 100.0 ng/mL). The amount of free PSA in serum samples was measured using an investigational double antibody immunoradiometric assay available through Dianon Systems, Stratford, Connecticut.” This assay recognizes only the free, uncomplexed form of PSA. Assay components consisted of a monoclonal anti-free PSA antibody (Immunocorp Sciences, Montreal, Canada) coated onto polystyrene 12 X 75 mm tubes; affinity purified, rabbit polyclonal anti-PSA labeled with lz51 and free PSA standards created from human seminal plasma.‘* The mass of free PSA was confirmed by both TOSOH and Yang PSA methods. Standards were prepared by the addition of a known mass of human seminal plasma-free PSA to a protein matrix buffer consisting of 0.01 M phosphate-buffered saline, 0.25% bovine serum albumin, and 0.01% sodium azide (NaN,). Assays were performed at 23°C. One hundred microliter of neat serum and 100 pL of assay buffer were added to the coated tubes and incubated for 2 hours with constant low-speed shaking. Tubes were washed two times with 4 mL of wash buffer (0.01 M phosphate buffered saline, 0.25% bovine serum albumin, 0.01% Tween 20, and 0.01% NaN,) followed by the addition of 200 pL of lz51 anti-PSA. After 2 hours of incubation with low-speed shaking, tubes were twice washed with 4 mL of wash buffer, decanted, and the total bound radioactivity determined in a gamma counter. The amount of free PSA specifically captured by the solidphase monoclonal anti-free PSA was estimated by extrapolation of the amount of 1251 tracer bound against a standard curve. A free PSA standard curve representing two separate lots of iodinated tracer is illustrated in Figure 2. Specificity was demonstrated by reactivity profile analysis of fractionated patient specimens. Male serum samples with known total PSA values were chromatographed by high-performance liquid chromatography on a Superose-6 I.Ea.oGY@
46 (2), 1995
0.1. F-PSA
Dilution
(ng/mL)
Typical free prostate-specific antigen (FPSA) calibration curve. Dynamic range of 0.0 to 50.0 ng/mL. Plotted are the mean f 2 SD of each standard from four separate runs and from two separate tracer lots (lot l[..V...], lot 2 [-*-I). B/T% is the percentage of bound counts per minute (CPM) over the total CPM added. FIGURE 2.
(Pharmacia, Piscataway, NJ) support under nondenaturing conditions. Resultant fractions were analyzed for total and free PSA. The total PSA reactivity profile of these fractionated serum specimens clearly showed two discrete peaks of activity: the first at approximately 90 kd corresponding to the PSA-ACT complex and the second at approximately 30 kd representing the free PSA (Fig. 1). The free PSA assay’s selectivity for free PSA was demonstrated repeatedly in both native serum specimens and also with artificially prepared samples of total PSA containing different proportions of free PSA to PSA-ACT (data not shown).
TABLE
Sample Nl*
N2 N3 N4 N5 Pl P2 P3 P4 P5 Mean percent recovery*
1 3.92* 0.16 0.25 0.37 0.10 6.45 0.43 0.69 2.16 1.27
10
1
I.
Freeze/thaw
Factor
FIGURE 3. Serial dilution curves of three cancer patient serum samples. Five dilutions for each sample. The mean value derived from the average of four replicates. Sample 1 b, mean observed values; - -, expected dilution line) y = 1. lx - 0.19, r = 0.99. Sample 2 (m,mean observed values; ...... expected dilution line) y = 1.0 lx + 0.59, r = 0.99. Sample 3 (A, mean observed values;-..-, expected dilution line) y = 1.18x + 0.32, r = 0.99. The free PSA assay demonstrated an analytical sensitivity range of 0.02 to 0.04 ng/mL, which was determined by comparing the mean of the lowest standard (0.2 ng/mL) and the 99% confidence interval of the assay’s zero calibrator over 10 replicates performed on 3 different days. The free PSA assay possesses an acceptably low total coefficient of variation, 10% or less, for combined inter- and intra-run variability across the measuring range of the assay. The assay was linear across its measuring range (0.0 - 50.0 ng/mL) (Fig. 3). Free PSA levels in specimens were not affected by up to 10 repeated minus 70” to 23°C freeze/thaw cycles (Table I) and
stability
Cycle Number 2 6 4.39 4.56 0.22 0.22 0.25 0.29 0.38 0.41 0.11 0.16 6.14 6.31 0.43 0.45 0.71 0.70 2.15 2.07 1.30 1.24 106 114
of the free PSA antigen 10 4.53 0.20 0.32 0.41 0.13 7.07 0.45 0.74 2.08 1.29 113
Overall Mean 4.35 0.20 0.28 0.39 0.13 6.49 0.44 0.7 1 2.12 1.28
cv (%) 6.8 14.1 12.3 5.3 21.2 6.2 2.6 3.0 2.2 2.1
KEY: CV = coefficient of variation. ‘Sample type: N = normal sewn samples; P = prostate cancer patient serum samples. ‘Values represent the average of4 replicate determinations offree PSA in ng/mL. *Determined by the comparison ofeach additional cycle tocycle I.
UROLOGYa 46 (2), 1995
189
TABLE II. Sample Nl* N2 N3 N4 N5 Pl P2 P3 P4 P5 Mean percent recovery*
Day-to-day
stability
of tee free PSA antigen
Cycle Number 3 2 3.88 4.09 0.18 0.22 0.25 0.24 0.40 0.44 0.08 0.12 9.53 9.58 2.25 2.21 0.46 0.46 1.91 1.88 0.92 0.89 104 105
1 3.92+ 0.16 0.25 0.37 0.10 9.65 2.30 0.39 1.95 0.90
6 4.32 0.22 0.3 1 0.30 0.16 9.22 2.15 0.54 1.97 0.92 114
Overall Mean 4.05 0.20 0.26 0.38 0.12 9.50 2.23 0.46 1.93 0.91
cv % 4.9 15.4 12.2 15.7 29.7 2.0 2.8 13.3 2.1 1.7
KEY: CV = coefficient of variation. *Sample type: N = normal serum samples; P = prostate cancer patient senm samples. ‘Values represent the average of4 replicate determinations offree PSA in ng/mL. *Determined by the comparison ofeach additional cycle to cycle I.
TABLE III.
Sample No. 1 2 3 4
Evaluation
Control FPSA (n&W 0.43 0.96 1.41 8.37
of the effect of intetiering Hemoglobin (500 mgldl) FPSA (w/mL) 0.39 1.00 1.40 8.20
Recovery* WI 90.7 104.2 99.3 98.0
substances
in the free PSA immunoassay
Bilirubin (20 mg/dL) FPSA Recovery WmLl (%I 0.38 88.4 1.02 106.3 1.37 97.2 100.0 8.37
(1000 FPSA (n&W 0.38 1.03 1.43 8.25
Lipid mg/dL] Recovery (%I 88.4 107.3 101.4 98.6
KEY: FPSA =free PSA. *Recovery determined with assay buffer as the control.
exhibited good room temperature stability over six repetitive temperature cycles between 23°C for 30 minutes followed by 24 hours at 4’C (Table II). Interfering substances were evaluated by the addition of known amount of interferents to serum pools containing known amount of free PSA. Controls were corrected for the volume effect. Table III demonstrates that none of the potential interferent substances (hemoglobin 500 mg/dL, bilirubin 20 mg/dL, lipid 1000 mg/dL) interfered with the accurate determination of free PSA.
DATA COLLECTION AND DATA ANALYSIS Total and free PSA tests were performed at Dianon Systems, with operators blinded to the clinical specimen type (cancer, benign, control). The proportion of free PSA in a patient’s sample was calculated by dividing the patient’s free PSA value by the total PSA value. This proportion, expressed as a percent, will be referred to as free/total PSA throughout the remainder of this article. A structured data abstract form was designed to obtain clinical information on the patient’s age, history of prostate disease, history of prostate-related treatment, history of cancer, and needle biopsy results. Data were abstracted from the computer database in each of the three specimen banks. A nonparametric procedure (Wilcoxon) was used to test the statistical significance of differences in the median values of total PSA and free/total PSA between BPH and prostate cancer patients. Sensitivity and specificity were calculated. Receiver-operating characteristic (ROC) curves were generated by plotting sensitivity versus specificity to evaluate the assay performance for total PSA and for free/total PSA. 190
RESULTS Frequency distributions of patient age at disease diagnosis, total PSA levels, and the free/total PSA levels are presented in Table IV. The majority of the patients, 68% with cancer, 72% with BPH, and 52% asymptomatic normal, were between the ages of 55 and 64 years. The mean age was 64 years for cancer patients, 62 years for men with BPH, and 62 years for asymptomatic normal control subjects. Distribution of total PSA levels was notably different among the three patient groups. All asymptomatic normal controls had total PSA levels below 4.0 ng/mL; 23 (37%) patients with BPH were below 4.0 ng/mL, and only 4 (7%) cancer patients had a total PSA below 4.0 ng/mL. More cancer patients (73%) had low levels of free/total PSA (less than 15%) than patients with BPH (29%) or normal control subjects (24%). Among cancer patients, the distribution of the Gleason score was as follows: 58% (32 of 55) Gleason 5 or 6, 24% (13 of 55) Gleason 7 or 8, 11% (6 of 55) Gleason 9 or 10, and 7% (4 of 55) Gleason score not available. Information on clinical stage, based on TNM stage grouping, was available for 71% (39 of 55) of the cancer patients. URoLmYm 46 (2), 1995
Frequency distributions of patient age at disease diagnosis, total PSA * and proportion of free to total PSA by patient group
TABLE
IV.
Patient Cancer No. (%) 55 (1001
Total samples Age (yrl 50-54 55-59 60-64 65-69 70-75 Total PSA (ng/mL] <4 4-10 >lO Free/total PSA (Oh) 510 10.1-15 15.1-20 20.1-25 25.1-30 >30
1 7 30 9 8
(2) (131 (55)
(161 (151
Group
Benign No. (%) 62 (100)
Normal Control No. (%) 64 (1 OO]
4 17 28 5 8
11 10 23 13 7
(61 (27) (45) (8) (131
(17) (16) (36) (201 (1 11
4 (71 25 (45) 26 (47)
23 (37) 32 (52) 7 (1 1)
64 (100) 0 (01 0 (0)
19 (35) 21 (38) 8 (151
7 11 12 11 7 14
3 12 10 16 5 18
6 (1 11 1 (2) 0 (01
(111 (18) (19) (18) (1 11 (23)
(51 (19) (16) (25) (8) (28)
* TOSOH immunoassay
Among these patients, 11 were Stage I, 21 were Stage II, 3 were Stage III, and 4 were Stage IV Descriptive statistics for total PSA and the free/total PSA are summarized in Tables V and VI. When all subjects were included in the analysis (Table V), the median value for total PSA was highest among cancer patients (9.60 ng/mL), followed by patients with BPH (4.89 ng/mL), and was lowest among normal control subjects (1.07 ng/mL). A nonparametric Wilcoxon test indicated that the
differences in the medians of total PSA between cancer and BPH patients were statistically significant (P
Descriptive statistics for total PSA * and proportion free to total PSA for all subjects by patient group
TABLE V.
Patient Normal Control (n = 64) Total PSA (ng/mL) Mean Median Standard deviation 95% Cl for mean Free/Total PSA (%) Mean Median Standard deviation 95% Cl for mean KEY: 95% CI = 95 percent * TOSOH immunoassay. ‘Wilcoxon test comparing
confidence median
UROLOGYa 46 (2), 199.5
1.24 1.07 0.81 1.03- 1.44 26 23 14 23-30 interval values
Benign (n = 62)
of
Group Cancer (n = 55)
5.52 4.89 4.12 4.47-6.75
13.64 9.60 15.18 9.54-l 7.75
23 21 13 20-27
13 13 6 11-14
P Value’
< 0.000 1
< 0.000 1
of the mean.
between
benign
and cancer patients.
191
TABLE VI.
Descriptive
statistics for total PSA* and proportion of free to total PSA for subjects with total PSA 4.0 to 10.0 ng/mL by patient group Patient Croup Cancer (n = 25)
Benign (n = 32) Total PSA (ng/mL) Mean Median Standard deviation 95% Cl for mean Free/Total PSA (%) Mean Median Standard deviation 95% Cl for mean
6.18 6.05 1.48 5.65-6.72
P Value+
6.92 6.90 1.75 6.20-7.65
0.13
13 14 6 11-16
22 21 9 18-25
0.0004
KEY: 95% CI = 95 percentconfidence intervalfor the mean. * TOSOH immunoassay. tWilcoxon test comparing median values between benign and cancer patients.
100
80
0-l 0
FIGURE analysis free and primary disease.
/ 20
40 l-specificity
60
80
100
4. Receiver-operating characteristic curve comparing total PSA and the proportion of total (F/T) PSA based on 55 patients with prostate cancer and 62 patients with benign All PSA values are included.
clinical database, analysis revealed an improved performance for free/total PSA compared to total PSA (Fig. 5; Table VII). Free/total PSA had a clearly higher specificity than total PSA for almost all sensitivity levels in the total PSA range of 4.0 to 10.0 ng/mL (Fig. 5; Table VII). This observation was more profound when the value of free/total PSA was higher. At free/total PSA values 25% or less, all cancers were detected, and 31% of the patients with BPH were correctly identified. This selectivity is in contrast to total PSA 4.0 ng/mL or higher, which correctly identified all cancer patients but could not identify any patients with BPH (Table VII). COMMENT
The objective uate the clinical 192
of the current study was to evalutility of free PSA to differentiate
0, : I ;-0
20
40 I -epecif$
1 : : IA 80
100
FIGURE 5. Receiver-operating characteristic (ROC) curve analysis comparing total PSA and the proportion of free and total (F/r) PSA based on 25 patients with primary prostate cancer and 32 with benign disease with total PSA values between 4 and 10 ng/mL.
between benign and malignant prostate disease in the diagnostic gray zone of total PSA values 4.0 to 10.0 ng/mL. Similar to Christensson et al.,7 we observed an overall significantly smaller proportion of free PSA in prostate cancer patients than in patients with BPH. More importantly, when data were limited to a range of 4.0 to 10.0 ng/mL total PSA, differences in the proportions of free to total PSA between cancer and benign patients remained statistically significant, whereas differences in total PSA values were not detectable (Table VI). The PSA test has become one of the most costeffective tools for detecting early prostate cancer. 2~13~14In general, a diagnosis of prostate cancer is uncommon at serum levels of total PSA at or below 4 ng/mL and is common at levels above 10 ng/mL. The diagnostic gray zone is between 4.0 and 10.0 ng/mL, where the differential diagnosis uROLoGYrn 46 (21, 1995
TABLE VII.
Sensitivity
and specificity for different total PSA and proportion values for subjects with total PSA 4.0 to 10.0 ng/mL
Sensitivity
[%)
(NoJtotal)
Specificity
of free to total PSA (%)
(NoJtotal)
Total PSA
(w/m Ll 24 25 26 27 >8 29 Free/Total PSA (%) 530 525 520 515 514 112 510 57
100.0 92.0 68.0 48.0 36.0 16.0
(25/25) (23/25) (17125) (12125) (9/25) (4/25)
0.0 28.1 50.0 65.6 90.6 93.8
(O/32) (9/32) (16132) (21/32) (29/32] (30132)
100.0 100.0 88.0 72.0 68.0 36.0 24.0 20.0
(25/25) (25/25) (22/25) (18125) (17125) (9/25) (6125) 15/251
15.6 31.3 50.0 75.0 81.3 81.3 87.5 100.0
(5/32) (10/32) (16/32) (24/32) (26/32) (26/32) (28/32) [32/32)
of prostate cancer is most difficult2J4 Our data showed similar results to prior studies in that there were significant differences in total PSA distribution between patient groups over all values of total PSA, but not in the range between 4.0 and 10.0 ng/mL (Tables V VI). Alternative methods for increasing the sensitivity and specificity of total PSA within this diagnostic gray zone have been proposed such as PSA density,15 PSA velocity,16 and age-specific reference range. l7 PSA density and PSA velocity have contributed significantly to disease understanding but have suffered practical limitation due to the difficulty of accurately measuring prostate volume for PSA density calculations and the inherent biologic and laboratory variability associated with serial PSA determinations required for PSA velocity calculations1*~r9 Agespecific reference ranges should increase the sensitivity of PSA in younger men and the specificity of PSA in older men.20 Our study results suggest that the proportion of free to total PSA has intrinsic value as an aid in the differential diagnosis of benign disease from cancer of the prostate in the range of 4.0 to 10.0 ng/mL total PSA (Table VI; Fig. 5). Based on our analysis, free/total PSA as a single test did not differentiate between benign and prostate cancer patients significantly better than the total PSA alone over the entire PSA range (Fig. 4). If free/total were to be used as a single test, the optimal sensitivity and specificity would be 87% and 52%, respectively, at a cutoff of 20% free/total PSA based on the current data set. Compared with a total PSA cutoff of 4 ng/mL, free/total PSA improved the specificity from 37% to 52%; however, sensitivity worsened from 93% to 87%. As a sinUROLOGY@ 46 121, 1995
y/ /-l
4
Eknign / I 2l
Ii
40% Percent
60% Free /Total
80%
100%
PSA
FIGURE 6. Theoretical distributions for the proportion of free to total PSA in patients with prostate cancer and benign histologic results.
gle test, the sensitivity and specificity of a free/total PSA cutoff of 20% was essentially identical to a total PSA cutoff of 5 ng/mL. Theoretical distributions for free/total PSA values in subjects with total PSA values in the range of 4.0 to 10.0 ng/mL were constructed from our database (Fig. 6). These distributions were generated using curve fitting software.21 (To determine these distributions, we first fitted curves to our empirical cumulative data in each patient group. This process generated many different equations that fitted the data. Among all equations, a functional form known as the log-normal cumulative function was chosen because the equation had a very good fit [r2 >0.951 in both patient groups, and the functional equation is biologically plausible. That is, the shape of each distribution is similar to other biologically controlled measures such as blood pressure and 193
hematocrit. Once we identified the functional form of the distribution, the probability density function for each patient group can be easily derived. It is the probability density function that is plotted in Figure 4. The units on the vertical axis are arbitrary. They are derived from the fact that the total area under each curve must equal unity.) At a cutoff of 20% free/total PSA, the theoretical curves show a sensitivity of 80% and a specificity of 49%. These values are within the 95% confidence interval of our data estimate of a sensitivity of 88% and a specificity of 50% at the same free/total PSA cutoff. The figure clearly shows the effective separation that can be attributed to the proportion of free/total PSA in the diagnostic gray zone of total PSA. One of the most interesting observations in our total PSA 4 to 10 ng/mL subdata set was the absence of prostate cancer in patients with free/total PSA more than 25%. Thirty percent of all patients with benign conditions whose original total PSA values lay between 4.0 and 10.0 ng/mL fell into this category (Table VII). Age is an important factor affecting prostate gland volume and the prostate’s ability to produce and secrete PSA.1° Both PSA level and the volume of the prostate gland increase with increasing age. lo In the current study, we were able to reduce the potential confounding effect of age by restricting patient’s age to the range of 50 to 75 years. The frequency distribution of age was similar among the three patient groups, although the mean age was slightly greater in prostate cancer patients (64 years) than in patients with BPH, or healthy normal control subjects (62 years) (Table IV). Misclassification of patient status, particularly patients with benign prostate disease, was unlikely in the current study because sextant biopsies were performed for patient disease confirmation. REFERENCES 1. Wang MC, Valenzuela LA, Murphy GP, and Chu TM: Purification of a human prostate-specific antigen. Invest Urol 17: 159-163, 1979. 2. Stamey TA, Yang N, Hay AR, McNeil JE, Freiha FS, and Redwine E: Prostate-specific antigen as a marker for adenocarcinoma of the prostate. N Engl J Med 317: 909-916, 1987. 3. Christensson A, Laurel1 CB, and Lilja H: Enzymatic activity of the prostate-specific antigen and its reactions with extra-cellular serine proteinase inhibitors. Eur J Biochem 194: 755-763, 1990. 4. Lilja H, Christensson A, Dahlen U, Matikainen M, Nilsson 0, Petterson K, and Lovgren T: Prostate-specific antigen in serum occurs predominantly in complex with a,antichymotrypsin. Clin Chem 37: 1618-1625, 1991. 5. Stenman UH, Leinonen J, Alfthan H, Rannikko S, Tuhkanen K, and Alfthan 0: A complex between prostate-
specific antigen and a,-antichymotrypsin is the major form of prostate-specific antigen in serum of patients with prostatic cancer: assay of the complex improves clinical sensitivity for cancer. Cancer Res 51: 222-226, 1991. 6. Partin AW, and Oesterling JE: The clinical usefulness of prostate specific antigen: update 1994. J Urol 152: 1358-1368, 1994. 7. Christensson A, Bjork T, Nilsson 0, Dahlen U, Matikainen MT, Cockett AT, Abrahamsson PA, and Lilja H: Serum prostate-specific antigen complexed to ai-antichymotrypsin as an indicator of prostate cancer. J Urol 150: 100-105, 1993. 8. Wood WG, van der Sloot E, and Bohle A: The establishment and evaluation of luminescent-labelled immunometric assays for prostate-specific antigen-a,-antichymotrypsin complexes in serum. Eur J Chem Clin Biochem 29: 787-794, 1991. 9. Leinonen J, Lovgren T, Vornanen T, and Stenman UH: Double-label time-resolved immunofluorometric assay of prostate-specific antigen and of its complex with c~iantichymotrypsin. Clin Chem 39: 2098-2103, 1993. 10. Oesterling JE, Cooner WH, Jacobsen SJ, Guess HA, and Lieber MM: Influence of patient age on the serum PSA concentration. An important clinical observation. Urol Clin North Am 20: 671-680, 1993. 11. Cuny C, Pham L, Kramp W, Sharp T, and Soriano TF: Evaluation of a two-site immunoradiometric assay for the measurement of non-complexed (free) PSA. Clin Chem 1995. (submitted) 12. Rusciano D, Berardi A, Ceccarini C, and Terrana B: Concomitant purification of prostatic carcinoma tumor markers from human seminal fluid under nondenaturing conditions Clin Chem 34: 2528-2532, 1988. 13. Brawer MK, Chetner MP, Beatie J, Buchner DM, Vessella RL, and Lange PH: Screening for prostatic carcinoma with prostate specific antigen. J Urol 147: 841-845, 1992. 14. Catalona WJ, Smith DS, Ratliff TL, Dodds KM, Coplen DE, Yuan JJ, Petros JA, and Andriole GL: Measurement of prostate-specific antigen in serum as a screening test for prostate cancer. N Engl J Med 324: 1156-1161, 1991. 15. Benson MC, and Olsson CA: Prostate specific antigen and prostate specific antigen density. Roles in patient evaluation and management. Cancer 74: 1667-1673, 1994. 16. Carter HB, Pearson JD, Metter EJ, Brant LJ, Chan DW, Andres R, Fozard JL, and Walsh PC: Longitudinal evaluation of prostate specific antigen levels in men with and without prostate disease. JAMA 267: 2215-2220, 1992. 17. Oesterling JE, Jacobsen SJ, Chute CG, Guess HA, Girman CJ, Panser LA, and Lieber MM: Serum prostate-specific antigen in a community-based population of healthy men. Establishment of age-specific reference ranges. JAMA 270: 860-864, 1993. 18. Mettlin C, Littrup PJ, Kane RA, Murphy GP, Lee F, Chesley A, Badalament R, and Mostofi FK: Relative sensitivity and specificity of serum prostate specific antigen (PSA) level compared with age-referenced PSA, PSA density, and PSA change. Cancer 74: 1615-1620, 1994. 19. Stamey TA: Second Stanford Conference on International Standardization of Prostate-Specific Antigen Immunoassays: September 1 and 2, 1994. Urology 45: 173-184, 1995. 20. Catalona WJ, Hudson MA, Scardino PT, Richie JP, Ahmann FR, Flanigan RC, deKernion JB, Ratliff TL, Kavoussi LR, Dalkin BL, et al: Selection of optimal prostate specific antigen cutoffs for early detection of prostate cancer: receiver operating characteristic curves. J Urol 152: 2037-2042, 1994. 21. Jandel Scientific: TableCurve 20 for Windows. AISN Software Inc., San Raphael, California, 1994.
UROLOGY’
46 (21, I995