PROSTATE SPECIFIC ANTIGEN DENSITY OF THE TRANSITION ZONE FOR EARLY DETECTION OF PROSTATE CANCER

0022-5347/98/1602-0411$03.00/0

THEJOURNAL OF UROLOGY

Vol. 160,411-419,August 1998 Printed in U S A

Copyright 0 1998 by AMERICAN UROLOGICAL ASSOCIATION, INC.

PROSTATE SPECIFIC ANTIGEN DENSITY OF THE T W S I T I O N ZONE FOR EARLY DETECTION OF PROSTATE CANCER B. DJAVAN, A. R. ZLOTTA,* G. BYTTEBIER, S. SHARIAT, M. OMAR, C. C. SCHULMAN M. MARBERGER

AND

From the Departments of Urology, University Hospital of Vienna, Vienna, Austria, and Erasme Hospital, University Clinics of Brussels, Brussels, Belgium

ABSTRACT

Purpose: We compare the ability of several prostate specific antigen (PSA) parameters, including PSA density of the whole prostate and of the transition zone, percent free PSA and PSA velocity, t o enhance the specificity for prostate cancer detection and to reduce unnecessary biopsies in men with serum PSA levels of 4 to 10.0 ng./ml. Materials and Methods: This prospective study included 559 consecutive men referred for early prostate detection or lower urinary tract symptoms who had a serum PSA of 4.0 to 10.0 ng./ml. All men underwent prostatic ultrasonography and sextant biopsy with 2 additional transition zone biopsies. Specific biopsies of abnormal findings on digital rectal examination were also performed. In all cases if first biopsies were negative an additional set of biopsies was performed within 6 weeks. The ability of PSA density, PSA transition zone, PSA velocity and percent free PSA to improve the power of PSA in the detection of prostate cancer was evaluated with univariate and multivariate analyses as well as receiver operating characteristics (ROC) curves. Results: Of 559 patients 342 had histologically confirmed benign prostatic hyperplasia and 217 had prostate cancer. Mean serum PSA, PSA velocity, PSA density and PSA transition zone were statistically higher (p <0.018,p <0.037,p <0.0001 and p <0.0001,respectively) and percent free PSA was statistically lower (p <0.0001)in patients with prostate cancer than in those with benign disease. Multivariate analysis and ROC curves showed that PSA transition zone and percent free PSA were the most powerful and highly significant predictors of prostate cancer. Areas under the ROC curve for PSA transition zone and percent free PSA were 0.827 and 0.778, respectively (p = 0.01 McNemar test). Combination of free-to-total PSA with PSA transition zone significantly increased the area under the ROC curve compared to PSA transition zone alone (p = 0.020).With a 95% sensitivity for prostate cancer detection a PSA transition zone cutoff of 0.25 ng./ml./cc would result in the lowest number of unnecessary biopsies (47% PSA transition zone specificity) compared to all other PSA parameters. However, total prostate volume (greater than 30 cc in 422 men or less than 30 cc in 137) was an important factor in predicting the statistical performance of PSA transition zone. In fact, PSA transition zone did not outperform free percent PSA in sensitivity and specificity when the entire prostate gland volume was less than 30 cc (p = 0.094 McNemar test). Conclusions: PSA density of the transition zone enhances the specificity of serum PSA for prostate cancer detection in referred patients with a serum PSA of 4.0to 10.0 ng./ml. compared to other PSA parameters currently available. While PSA transition zone was more effective in prostates greater than 30 cc and percent free PSA was more effective in prostates less than 30 cc, the combination of percent free PSA with PSA transition zone further increased prostate cancer prediction. KEYWORDS:prostate-specific antigen, prostatic neoplasms Prostate cancer has become the first neoplasm in men and the second leading cause of cancer death.l.2 Attempts to reduce the mortality mainly focused on early detection of this disease. Since its discovery in 1979, prostate specific antigen (PSA) has unequivocally proved its usefulness as a serum marker for prostate cancer. However, in patients with a PSA below 10 ng./ml. an important overlap exists between benign prostatic hyperplasia (BPH) and prostate cancer. Indeed, PSA is not prostate cancer specific and develops at an age when the prevalence of BPH is high. Previous reports indicate that two-thirds of the patients who undergo biopsies based on a PSA of 4 to 10 ng./ml. have no histological eviAcce ted for publication March 13, 1998. Reacfat annual meeting of American Urological Association, New Orleans, Louisiana, April 12-17, 1997. * Supported by a grant from the Erasme Foundation.

dence of prostate ~ a n c e rIn . ~this PSA range several concepts have been introduced, all aiming to optimize the clinical use of PSA by improving its sensitivity and specificity, and trying to decrease the number of unnecessary biopsies in men with benign disease. These concepts include PSA density (relating the serum PSA to the volume of the prostate), PSA velocity (evaluating the rate of change of PSA values with time), age adjusted PSA reference ranges (adjusting the PSA level with patient age) and determination of PSA molecular forms (free versus protein bound molecular forms of PSA). The clinical usefulness of free PSA measurement in men with intermedibut - ~the value of the ate PSA levels has been d e m ~ n s t r a t e d ~ other PSA parameters for enhancing prostate cancer detection specificity compared to PSA alone is still the subject of debate.3.*-11

411

412

PROSTATE SPECIFIC ANTIGEN DENSITY OF TRANSITION ZONE

Recently, a new parameter, PSA density of the transition zone, which relates the serum PSA level to the volume of the transition zone of the prostate, has been investigated. 12-14 This concept derives from the observation that in the benign prostate most PSA leaking into the serum comes from the transition zone.15 BPH results almost exclusively from hyperplasia of the transition zone and, therefore, peripheral gland production of PSA is relatively constant as the gland enlarges due to BPH.15-17 We recently demonstrated in a retrospective study in men with PSA below 10 ngJml. and histologically proved benign or malignant disease that PSA transition zone was more accurate than PSA density of the total prostate for detecting prostate cancer.14 To evaluate further the usefulness of this new parameter for prostate cancer prediction in men with intermediate serum levels, a multicenter prospective study was initiated comparing PSA transition zone with PSA density, age adjusted PSA, PSA velocity and free-to-total PSA ratio. Our goal was to determine the most accurate parameter or combination of PSA parameters which would significantly enhance the detection of prostate cancer while at the same time minimize the number of unnecessary prostate biopsies in patients with benign disease. MATERIALS AND METHODS

same test were included in the study. The increase in PSA was given in ngJml. per year. Deep frozen serum ( - 7 0 0 was used to determine total serum PSA and free-to-total serum PSA ratio using the widely evaluated, Delfia Pro Status? PSA fredtotal assay.7 A nonradioactive agent (europium) is used, which gives an equimolar detection of PSA in free and complexed forms. Correlation coefficients between the commercial assay and other currently available tests (Hybritech Tandem E) have been demonstrated to be at 0.99.7 Mean intra-assay variation coefficient with this method in our study was 1.8% and interassay variation coefficient averaged 2.7%. All sample determinations were run in duplicate. Digital rectal examination was categorized as normal or suspicious for cancer. Abnormal digital rectal examination was defined as any examination that was suspicious for cancer by the examining urologist. Glands that were simply enlarged were not classified as abnormal for the purposes of the study. Histopathological findings were classified as negative (normal or BPH) or positive (cancer). A Gleason score was used to grade tumors. Statistical analysis. The observed 5th, 25th, 50th (median), 75th and 95th percentiles were calculated on the basis of the empirical distribution of the data in the entire population and in each 10-year age group. For continuous data groups were compared by Wilcoxon rank sum test. Continuous variables were assessed using univariate and multivariate logistic regression analyses. For each test, separately in each decade and according to prostate size, sensitivity and specificity of the analyzed parameters were calculated. Sensitivity was defined as equal to the number of true positives divided by the sum of true positives plus false-negative. Specificity was defined as equal to the number of true negatives divided by the sum of the true negatives plus false-positives. Positive and negative predictive values were calculated as well. Receiver operating characteristics (ROC) curves in which the value for sensitivity was plotted against the false-positive rate (1 - specificity) were generated for each test. The closer the ROC curve is to the upper left of the graph and the larger the area under the curve, the better the test. Statistical analysis of the different areas under the curve was calculated with the McNemar test, and p c0.05 was considered statistically significant. Data were analyzed with computer software.

Patients. From January 1996 to January 1997, 559 patients 49 to 82 years old were enrolled in a prospective study at the Departments of Urology of the University of Vienna, Austria and of the Erasme Hospital, University Clinics of Brussels, Belgium. The study population consisted of men referred for either early cancer detection or lower urinary tract symptoms. All men with a serum PSA of 4 to 10 ng./ml. were included in the study. Patients with a history of prostate cancer, acute or chronic prostatitis and histological diagnosis of prostatic intraepithelial neoplasia of any grade, or those were in urinary retention, with an indwelling catheter or with proved tract urinary infection were excluded from study. Study design. All men underwent transrectal ultrasound with systematic sextant needle biopsies and 2 additional transition zone biopsies. Biopsies were also taken from suspicious areas at ultrasonography or digital rectal examination. In all cases if first biopsies were negative (benign), within 6 weeks a n additional set of sextant biopsies (and 2 transition zone biopsies) was routinely performed t o miniRESULTS mize sampling errors. Serum PSA determination was done before digital rectal Age, prostate and transition zone volume, serum PSA and examination or any endoscopic procedure with a radioimmu- PSA parameters (density, transition zone, velocity, free-tonoisometric assay.* Transrectal ultrasound was performed total ratio). Of the patients 342 had a benign histology and with a biplanar 7.5 MHz. probe. Total prostate and transition 217 had cancer diagnosed on biopsies. Table 1 shows the zone volumes were calculated using the prolate ellipsoid for- mean and standard deviation of age, serum PSA, total and mula (volume = 0.52 X length X width X height). The pros- transition zone volumes, free-to-total PSA ratio, PSA veloctate was scanned in the transverse and sagittal planes with ity, and PSA density of the total prostate and transition zone the subject in the left lateral decubitus position.14 The tri- in each group. Statistically significant differences included axial distances at the maximal diameter of the length, height larger total and transition zone volumes, and higher free-toand width of the total prostate and transition zone were total PSA ratio in patients with benign disease (p = 0.0001, measured as previously described.14 The transition zone, ' iWallac Oy,Finland. which when hyperplastic corresponds to the prostate adenoma, is usually separated from the central and peripheral zones by a distinct layer of fibrous tissue that is clearly TABLE1 visible on ultrasonography. PSA density of the total prostate and transition zone, exMean Benign Mean Prostate p Value Disease (SD) Ca (SD) ( t test) pressed as ng./ml.lcc, was calculated as the quotients of the serum PSA divided by either the total prostate volume or the No. pts. 342 217 0.1924 68.36 (8.62) 66.74 (7.54) transition zone volume, respectively. PSA velocity related to Age prostate vol. (cc) 48.57 (17.97) 35.50 (11.66) <0.0001 the rate of change in serum PSA values with time. PSA Total Transition zone vol. (cc) 25.72 (11.59) 14.77 (6.94) <0.0001 velocity was determined in patients who had had at least 3 Serum PSA (ngJml.) 6.72 (1.68) 7.46 (1.76) 10.018 PSA measurements each 12 months apart before the present Free-to-total PSA ratio (%) 31.84 (16.96) 16.09 (10.34) 10.0001 1.40 (1.17) 1.79 (1.45) 0.037 study. Only those patients who had PSA measured with the PSA velocity (ng./ml./yr.)

* Hybritech, Inc., San Diego, California.

PSA density (ng./ml./cc) PSA transition zone (ng./ml./cc)

0.156 (0.068) 0.316 (0.181)

0.231 (0.096) 0.621 (0.372)

10.0001 <0.0001

PROSTATE SPECIFIC ANTIGEN DENSITY OF TRANSITION ZONE

413

p = 0.0001 and p = 0.0001 respectively, Wilcoxon test). TABLE2. Sensitivity, specificity, and positive predictive and Serum PSA, PSA velocity, PSA density and PSA transition negative values at different cutoff values zone were statistically higher in patients with prostate canPos. Neg. cer (p = 0.018, p = 0.037, p = 0.0001 and p = 0.0001, %Sensitivity %Specificity Redictive Redietive value value respectively, Wilcoxon test). PSA transition zone values in patients with either prostate cancer or benign histology are Frpe-to-totol PSA and PSA 4.0-10 ng.I d . (521 p a . ) shown in figure 1 in the form of a box and whiskers plot % Free-tc-total PSA demonstrating the observed 5th, 25th, 50th (median), avercutof? 10 33.690 92.308 0.74118 0.68041 age, 75th and 95th percentiles. Cutoff values corresponding 0.62963 15 54.545 79.021 0.72669 to the values between P75 for cancer (75% of the cases are 20 68.984 69.930 0.60000 0.77519 above this cutoff value) and P25 for benign disease (75% of 30 87.701 54.196 0.55593 0.87079 the cases are equal to or below this cutoff value) for PSA 40 96.257 32.168 0.48128 0.92929 transition zone was 0.368 ng./ml./cc. For PSA density and PSA wlocuy and PSA 4.0-10 ng.Id.(307pts.) free percent PSA the cutoff values were 0.172 ng./ml./cc PSA velmity cutoff and 21.1%, respectively. (nghnuyr.): 0.25 2.326 97.753 0.42857 0.58oOo Sensitivity, specificity and positive predictive value of each 0.75 42.636 71.348 0.51887 0.63184 PSA parameter. Table 2 shows the sensitivity, specificity and 63.566 48.876 0.47399 1.25 0.64925 positive predictive value in diagnosing prostate cancer for 68.992 42.135 0.46354 0.65217 1.75 free-to-total PSA, PSA velocity, PSA density and PSA trans2.25 75.969 33.708 0.45370 0.65934 ition zone. When a 95th percentile was used as a cutoff to PSA &mi@ and PSA 4.0-10 ng.ld. (559pts.) maintain a 95% sensitivity, specificity for all PSA paramedensity cutoff ters was markedly reduced. However, as indicated in table 3, PSAbg/mvcc): a t 95% sensitivity for detecting prostate cancer specificity of 0.43816 0.10 96.313 21.637 0.90244 0.51964 0.15 79.263 53.509 0.80263 PSA transition zone was significantly better than all other 0.20 57.604 80.994 0.65789 0.75068 parameters (p = 0.0001 McNemar test). A PSA transition 0.25 33.641 89.474 0.66972 0.68oOo zone cutoff value of 0.25 ngJml./cc maintained a 95% sensiPSA tMnsifionzone and PS4 4.0-10 ng./mL (559pis.) tivity for detecting prostate cancer while sparing 47% of PSA transition zone unnecessary biopsies. cutoff ~ngJmvcc): Multivariate analysis. A multivariate logistic regression 0.15 100 7.602 0.407 1.ooO analysis model to predict prostate cancer applying stepwise 0.456 0.956 0.20 98.157 25.731 selection procedures based on 4 variables, after PSA velocity 0.532 0.25 94.931 47.076 0.936 0.30 88.018 61.988 0.595 0.890 had been excluded because of its low predictive value, is 0.35 79.724 72.515 0.647 0.849 shown in table 4. PSA density and PSA were further ex0.50 51.152 87.719 0.725 0.738 cluded because of a low predictive value. "he remaining

1.5; 1.4:

.*

1 1,2" 1 .lj 1 ,O" 0.9;

I

1.5000

0,8{ 0.70,621 749

0.6:

N B I

4

m

0,s: 0 4; I

!

0,3{ 0.20,l"

Od 1

Ca

B

FIG. 1. PSA density of transition zone (PSA-2'2). Data are expressed in nglmllcc in cases with prostate cancer (Ca)or benign ( B )disease. In each box plot lower and up r ends of whiskem represent 5th and 95th percentiles, and lower and upper ends of boxes represent 25th and 75th percentiles. Line insideEx represents median and central number is average.

414

PROSTATE SPECIFIC ANTIGEN DENSITY OF TRANSITION ZONE TABLE5.

TABLE3. Specificity for detecting prostate cancer for each of the PSA Dammeters amlvzed Variable

No, pts,

Cutoff value 95% Sensitivitv

% Corresponding

SWifiCitY

PSA (4-10) 559 4.4NgJml. 6.43 Velocity 307 4.1 NgJmlJyr. 7.87 Free-to-totalPSA 521 37% 36.02 PSA density 559 0.106 NgJmlJec 26.02 PSA transition zone 559 0.250 NgJmlJcc 47.08 The cutoff value which allows detection of 95% of the prostate cancers (95% sensitivity) is indicated for each parameter.

TABLE4. Multivariate logistic regression analysis model to predict ~ r ~ s t acancer te aonlvina steowise selection orocedure Logistic

Analysis Constant ( = intercept) PSA transition

viue Estimate

?

SE

Odds Ratio (95% CI)

-0.8465 2 0.3174 <0.001

5.2789 2 0.6446 196.152 (55.455-693.812)

PSA 4.0-10 NgJML Areaundertheeurve in ROCawesealCulatedanddram hreadldifferent p a r a m e t e r d Beparatelycnin combination(96):

PSA PSA velocity hbtlJtalPSA mdbensity

PSAtransitionzone combiition PSAkansition zone and k b t l J t a l PSA ~ m p a r i m nofthe d i f l i i t areasunder the curve using the McNemar test indicating the p valuea WO.05 staii3liCallysignificant)ofthe difference between different areas: PSA transition mne VB. PSA PSA transition zone vs. velocity PSA transition wne vs. htutotal PSA PSA transitionzone vs. PSA density combination PSA kansitim zone, h b t o t a l PSA VB. PSA laansitionmne h b t o t a l PSA vs. PSA density

61.2 58.0 77.8 76.2 82.7 88.1

0.001 0.001 0.027

0.011 0.020

0.206

znne

&%-total

PSA

<0.001

-0.0817 ? 0.0102

0.922

(0.903-0.940)

parameters, PSA transition zone and free-to-total PSA were the most powerful and highly significant predictors of prostate cancer. In this multivariate analysis PSA transition zone outperformed free-&total PSA for predicting prostate cancer (p = 0.001). ROC curves comparing PSA, P S A velocity, free-to-total PSA, P S A density and P S A transition zone. ROC curves

illustrate the sensitivity and specificity of each of the analyzed parameters. The closer the ROC curve approaches the upper left comer the greater the predictive power of the test. PSA transition zone was a significantly better predictor of prostate cancer than all other tests used (fig. 2). Indeed, as shown in table 5,area under the ROC curve for PSA transition zone was 82.7%, which is statistically larger than for all other tests investigated. Area under the curve was statistically larger for free-to-total PSA and PSA density than for

00 90 80

PSA-TZ -rc-

FTT PSA ,+*

70

60 50

PSAD

40

+

30 20

VEL

.-.B.-

PSA

++

I0

0

20

0 10

40

30

60 50

80 I00 70 90

100-Specificity (%) FIG.2. ROC curves of sensitivity plotted yainst 100 minus specificity in all atients with serum PSA in 4 to 10 ng./ml. PSA-2'2, PSA transition zone. FIT PSA, free-to-total PSA. SAD, PSA density. VEL,serum PgA and PSA velocity.

415

PROSTATE SPECIFIC ANTIGEN DENSITY OF TRANSITION ZONE

PSA but not for PSA density and free-to-total PSA (fig. 2). Combination of PSA transition zone and free-to-total PSA significantly increased the area under the curve compared to the use of PSA transition zone alone (p = 0.020 McNemar test). Znfluence of age, digital rectal examination and prostate size on PSA transition zone and percent free PSA sensitivity and specificity. Figure 3 shows the ROC curve for PSA trans-

ition zone according to age and digital rectal examination. Area under the curve was not statistically different whether patients were in the fBh, sixth or seventh decade, or whether they had a normal or abnormal digital rectal examination (p >0.05 McNemar test). Therefore, the statistical performance of PSA transition zone in predicting prostate caner was independent of digital rectal examination status. Similarly, digital rectal examination status had no impact on the statistical performance of PSA, PSA density, free-btotal PSA and PSA velocity. In fact, in patients with normal and abnormal digital rectal examination, respectively, area under the curve was 61 and 63.9%for PSA, 76.9 and 74%for PSA density, 83.6 and 80.1% for PSA transition zone, 78.5 and 75.6%for free-to-total PSA, and 58.2 and 55.7%for PSA velocity. Analyzing the predictive value of PSA transition zone and free-to-total PSA with respect to prostate size, figure 4 shows that PSA transition zone significantly lost its predictive power when the prostate was smaller than 30 cc (137 cases, area under the ROC curve 71.2%)compared to larger than 30 cc (422 cases, area under the ROC curve 83.5%,p = 0.001).In contrast, prostate size had no impact on the statistical per-

formance of free-to-total PSA. In prostates less than 30 cc area under the curve for PSA transition zone (71.2%)was comparable to the area for free-to-total PSA (74.0%, p >0.094 McNemar test). DISCUSSION

If one relies on PSA levels only to biopsy patients, a large number of unnecessary biopsies are performed in those with a benign prostate but moderately elevated PSA (less than 10 ngJml.). In our study we prospectively compared several methods (PSA density, PSA transition zone, free-to-total PSA, PSA velocity, age specific PSA reference range), all of which aim to improve the usefulness of PSA as a diagnostic tool by increasing its sensitivity and specificity. To minimize sampling errors we repeated sextant biopsies in all patients who had a benign histology on the first set of biopsies. Our study supports the effectiveness of PSA density of the transition zone as a new parameter for prostate cancer prediction. Indeed, in patients with a PSA of 4 to 10 ng./ml. the sensitivity, specificity and positive predictive value of PSA transition zone were significantly better than all other PSA parameters. The concept of PSA transition zone integrates the importance of the transition zone volume in interpreting PSA levels of patients without cancer. Above a PSA transition zone cutoff value it becomes unlikely that BPH accounts for the increase in PSA and, therefore, prostate cancer should be suspected. The previously calculated cutoff value of 0.35 ngJ

100 90 80 70 60 50 40 30 20

DRE N

+

ORE ABN

+

50-59yrs 43-

60-69yrs +K-

70-79yrs

+

10

o"~""""""l"'l 40

20

0

10

30

60

50

70

I00

80

90

100-Specificity (YO) FIG. 3. ROC c w e s of sensitivity plotted a g e t 100 minus specificity of PSA transition zone as fupction of 380 normal (DREN) or 179 ~ 239 in sixth and 190 in abnormal (DREABN) digital rectal examinations and as function of age. Curves are drawn for 79 patients L I fifth, seventh decades.

PROSTATE SPECIFIC ANTIGEN DENSITY OF TRANSITION ZONE

416

100 90 80 70 60 50 40 30 20

PSATZ <30 g

..*-

PSATZ>30 g _1

F/T < 309

.-eF/T > 309 -%

-

10

0

40

20 I0

30

60 50

80

70

100

90

'I00-Specificity (YO) FIG.4. ROC curves of sensitivity lotted against 100 minus specificity of PSA transition zone (PSA TZ) and percent free PSA ( F l T PSA) as function of whole prostate size ,?less than 30 (120cases) or greater than 30 (439) gm.

mlJcc for discriminating between patients with or without cancer14 was confirmed in this study and by others.18 In our initial work on the PSA transition zone concept we had purposely selected a patient population with more pronounced differences in PSA transition zone. Indeed, patients with benign disease had symptomatic BPH and, therefore, the transition zone was anticipated to be larger than in those with prostate cancer. Consequently, at comparable PSA levels PSA transition zone values were different in both groups. In the present study we prospectively included all patients referred either for lower urinary tract symptoms related to bladder outlet obstruction or for early prostate cancer detection. As indicated by its sensitivity and specificity PSA transition zone was less effective for predicting prostate cancer compared to our first study. With an area under the ROC curve of 0.82 PSA transition zone, nevertheless, clearly outperformed all other PSA parameters. The PSA transition zone cutoff value associated with the combination of the highest sensitivity (75%) and specificity (76%) was 0.35 ng./ ml./cc. However, missing 25% of prostate cancers is unacceptable in our clinical point of view. Therefore, when considering a 95% prostate cancer detection rate the PSA transition zone cutoff value in patients with a serum PSA of 4 to 10 ng./ml. should be 0.25 ngJmlJcc. In this PSA range maintaining a 95% sensitivity would result in 65 and 74% unnecessary biopsies when using free-to-total and PSA density, respectively, while this number would decrease significantly to 53% with PSA transition zone. Neither digital rectal examination status (normal or abnormal) nor patient age influenced PSA transition zone performance in predicting prostate cancer.

This finding appears logical, since PSA transition zone integrates transition zone volume and PSA levels, both of which have been demonstrated to increase in time with age. Therefore, the ratio is anticipated to remain stable if the increase is proportional.19 Others authors have also found that the use of PSA transition zone significantly enhances the prediction of prostate cancer in men with a serum PSA of 4 to 10 ng./m1.13.18.20 Creasy et a1 found that the use of PSA transition zone decreased the number of unnecessary biopsies in 240 consecutive men by 38% while 7.5% of malignant lesions would have been missed.18 Detecting 97% of cancers would have decreased the number of unnecessary biopsies by 22%. Horninger et a1 showed that while detecting 100% of malignant lesions, PSA transition zone avoided 18% of unnecessary biopsies.20 Maeda et al confirmed that PSA transition zone was more specific than PSA density in 94 patients from a community based urology practice with a normal digital rectal examination and 4 to 10 ng./ml. PSA.21 In contrast, Gohji et a1 studying Japanese men with 2.1 to 10 ng./ml. PSA, did not find an improvement for prostate cancer prediction with PSA transition zone compared to PSA density.Z2 Nevertheless, the conclusions of these authors were based on small series since 12 cancers only were detected in the 134 men in whom transition zone volume was measured. A limitation of the use of PSA transition zone is that prostate and transition zone size had a n impact in our prospective study. In fact, with a prostate smaller than 30 cc, in which the transition zone is not particularly enlarged and therefore more difficult to measure, the effectiveness of PSA

PROSTATE SPECIFIC ANTIGEN DENSITY OF TRANSITION ZONE

417

transition zone was significantly diminished. This finding is prostate size or PSA range.28-30 Stephan et a1 determined the comprehensible and implies that PSA transition zone is not ratio of free-to-total PSA to discriminate between prostate an ultimate test in screening populations when prostate sizes cancer and BPH, and showed that percent free PSA yielded are anticipated to be fairly small. Indeed, one should distin- significant results only in men with a prostate volume of less guish between patients recruited in screening programs and than 40 cc.27 Using the same test as Bangma et al but those referred to physicians for either lower urinary tract analyzing a different patient population, if areas under the symptoms or early detection of prostate cancer. When ana- curves were comparable (0.72versus 0.771,our free PSA ratio lyzing studies contradictory on the role of PSA density, it cutoff for detecting 90% of cancers in the 4.0 to 10 ng./ml. appears that a benefit was demonstrated usually in patients PSA range was 32%,whereas it was 22%in the Dutch study referred either for lower urinary tract symptoms or suspi- which included screening patients only. “his finding highcious digital rectal examination.9 Studies that failed to show lights the still unresolved problem of the percentage free PSA a benefit included patients from true screening programs.23 cutoff. In patients referred to the hospital for several reasons, inAnother mean to increase the accuracy of PSA as a screencluding lower urinary tract symptoms, and when prostate ing tool includes age specific reference ranges for serum PSA sizes are expected to be larger PSA transition zone definitely levels.11.17 In our study PSA transition zone clearly outperimproves prostate cancer prediction, avoids unnecessary bi- formed the use of the age specific PSA reference ranges for opsies and reduces cost. According to our data, the use of PSA predicting prostate cancer. The area under the curve for the transition zone is best for men with lower urinary tract PSA age specific reference range in the study of Morgan et al, symptoms, a PSA of 4 to 10 ng./ml. and a prostate larger than who analyzed 3,475 men with no clinical evidence of prostate 30 cc. According to previous recommendations on PSA, biopsy cancer and 1,783 men with prostate cancer, was significantly would be done on all of these patients.19 larger than what we found for PSA transition zone. Indeed, A possible problem associated with the use of PSA transi- the corresponding area under the curve for white men in the tion zone is related to the measurement of the zone. We 40, 50, 60 and 70-year age groups were 0.99, 0.98, 0.97 and reported that the transition zone volume measurement was 0.96, respectively, which made the age specific reference accurate and reproducible.24 However, all transrectal ultra- range PSA the nearly perfect test. However, again it should sound measurements are operator dependent and, therefore, be emphasized that the type of population as well the indisubject to possible variability. Furthermore, the meas- cations for biopsy were different, and one cannot extrapolate urement of the transition zone in small or large prostates such good results of age specific reference range to all patient sometimes might be difficult. Because transition zone volume populations. measurements are not routinely performed by urologists, The cost issue also must be considered. PSA transition clear agreements on such measurements should be obtained. zone or any prostate density measurement requires transrecAccording to our data using the commercial assay, the tal ultrasound equipment and, therefore, may be associated number of biopsies that would have been spared with free- with significant costs when compared to free-to-total PSA or to-total PSA ratio was similar to other rep0rts.5.~5*26In a PSA velocity. However, the additional cost created by the multicenter clinical trial evaluating free PSA in the differen- need for transrectal ultrasound equipment is offset by tiation of prostate cancer from benign disease in 622 screened the elimination of unnecessary biopsies and improved prosmen (301 with prostate cancer and 321 with benign disease) tate cancer prediction with PSA transition zone. with a 4.0 to 10 ng./ml. PSA percent free PSA decreased the number of unnecessary biopsies by 20 and 30% to detect 95 CONCLUSIONS and 90% of cancers, respectively.25 Analyzing 74 men with This prospective study supports our previous finding on lower urinary tract symptoms and a PSA of 4.1 to 10.0 the effectiveness of PSA transition zone for predicting prosng./ml., Morote et a1 found that a percent free PSA cut point tate cancer in patients with PSA of 4 to 10 ngJml. PSA of 25% detected at least 95% of prostate cancers and de- transition zone outperformed all other PSA parameters curcreased 26.9% of negative biopsies.5 rently available for predicting prostate cancer. In a referral Bangma et a1 studied a screening population of 1,726 men population with a PSA of 4 to 10 ng./ml. patients with a PSA in a retrospective analysis.7 Using the same free-to-total PSA transition zone of less than 0.25 could safely have been assay as in our study, they found that the application of spared unnecessary biopsy since less than 5% of cancers serum PSA of 4 ng./ml. or more, a free-to-total PSA ratio of would have been missed. For small prostates (less than 30 0.20 or less and an abnormal digital rectal examination as cc), which are more commonly found in screening populaindicators of biopsy would reduce the number of unnecessary tions, PSA transition zone is less effective then percent free biopsies by 37% but also reduce the number of detected PSA, and the use of the percent of free PSA is preferable. PSA cancer by l l % . 7 Vashi et al performed biopsies in 413 men transition zone should be applied to patients with larger with a percent free PSA of 0.24 or less and detected 95% of prostate volumes (greater than 30 cc) only. The combination the cancers but eliminated only 13%of the negative prostate of percent free PSA with PSA transition zone increased the biopsies.6 However, in contrast to PSA transition zone and prediction of prostate cancer. However, free PSA ratio cutoff earlier studies on free-to-total PSA, using a cutoff of 30 cc we determination is far from being definitively resolved and its could not detect any volume dependent difference in the clinical value in the decision to biopsy an individual patient performance of free-to-total PSA.4.27 Since in patients with a remains unclear. prostate smaller than 30 cc the effectiveness of the free PSA ratio for predicting prostate cancer is better than PSA transREFERENCES ition zone it would be logical to advocate, based on our re1. Parker, S. L., Tong, T., Bolden, S. and Wingo, P. A.: Cancer sults, the use of a test that requires venipuncture only. We statistics, 1996.CA, 46:5, 1996. found that the combination of PSA transition zone and free2. Dijkman, G. A. and Debruyne, F. M.: Epidemiology of prostate to-total PSA significantly increased prostate cancer prediccancer. Eur.Urol., 30: 281, 1996. tion and could further decrease the number of unnecessary 3. Brawer, M.11:Prostate-specificantigen: critical issues. Urology, biopsies. If both tests indicate a prostate cancer probability of 44:9,1994. less than 5% we might safely refrain from biopsy. 4. Catalona, W. J., Smith, D. S., Wolfed, R. L., Wang, T. J., However, all problems regarding free PSA percentage useRittenhouse, H. G., Ratliff, T. L. and Nadler, R. B.: Evaluation of percentage of free serum prostate-specific antigen to imfulness in clinical practice are far h m being resolved. Inprove specificity of prostate cancer screening.J.A.M.A.,274: deed, the best ratio of free-to-total PSA remains undeter1214. 1995. mined, and results may be influenced by assay type, Or

4 18

PROSTATE SPECIFIC ANTIGEN DENSITY O F TRANSITION ZONE

and Southwick, P. C.: Comparison of prostate specific antigen concentration versus prostate specific antigen density in the Encabo, G., de Torres, I. and Andreu, J.: Comparison of perearly detection of prostate cancer: receiver operating characcent free prostate specific antigen and prostate specific antiteristic curves. J. Urol., 152 2031, 1994. gen density as methods to enhance prostate specific antigen specificity in early prostate cancer detection in men with nor- 24. Zlotta, A. R., Djavan, B., Roumeguere, T., Marberger, M. and Schulman, C. C.: Transition zone volume on transrectal ultramal rectal examination and prostate specific antigen between sonography is more accurate and reproducible than the total 4.1 and 10 ngJml. J. Urol., 158. 502,1997. prostate volume. Brit. J. Urol., suppl., 80: A 926,1997. 6. Vashi, A. R., Wojno, K. J., Henricks, W., England, B. A., Vessella, R. L., Lange, P. H., Wright, G. L., Jr., Schellhammer, P. F., 25. Catalona, W. J., Partin, A. W., Slawin, K. M., Brawer, M. K., Patel, A., Flanigan, R. C., Riche, J . P., deKernion, J. B., Weigand, R. A., Olson, R. M., Dowell, B. L., Borden, K. K and Walsh, P. C., Scardino, P. T., Lange, P. H., Herschman, J. D., Oesterling, J. E.: Determination of the “reflex range” and appropriate cutpoints for percent free prostate-specific antigen Subong, E. N. P., Petteway, J. C., Parson, R. E., Loveland, in 413 men referred for prostatic evaluation using the AxSYM K G., Gasior, G. H. and Southwick, P. C.: A multicenter system. Urology, 49: 19,1997. clinical trial evaluation of free PSA in the differentiation of 7. Bangma, C. H., Rietbergen, J . B. W., Kranse, R., Blijenberg, prostate cancer from benign disease. J. Urol., part 2,167:111, B. G., Petterson, K and Schroder, F. H.: The free-to-total abstract 434, 1997. prostate specific antigen ratio improves the specificity of pros- 26. Chen, Y.T., Luderer, A. A,, Thiel, R. P., Carlson, G., Cuny, C. L. tate specific antigen in screening for prostate cancer in the and Soriano, T. F.: Using proportions of free to total prostategeneral population. J . Urol., 167: 2191, 1997. specific antigen, age and total prostate-specific antigen to pre8. Bedushi, M. C. and Oesterling, J. E.: Prostate-specific antigen dict the probability of prostate cancer. Urology, 47: 518,1996. density. Urol. Clin. N. Amer., 24: 323, 1997. 27 Stephan, C., Lein, M., J u g , K., Schnorr, D. and Loening, S. A,: 9. Bazinet, M., Meshref, A. W., Trudel, C., Aronson, S., PBloqUin, The influence of prostate volume on the ratio of free to total F., Nachabe, M., Begin, L. R. and Elhilali, M. M.: Prospective prostate specific antigen in serum of patients with prostate evaluation of prostate-specific antigen density and systematic carcinoma and benign prostate hyperplasia. Cancer, 7 9 104, biopsies for early detection of prostatic carcinoma. Urology, 43: 1997. 44,1994. 28 Roehrborn, C. G., Gregory, A., McConnell, J . D., Sagalowsky, 10. Carter, H. B. and Pearson, J . D.: Prostate-specific antigen velocA. I. and Wians, F. H., Jr.: Comparison of three assays for total ity and repeated measures of prostate-specific antigen. Urol. serum prostate-specific antigen and percentage of free Clin. N. Amer., 24: 333, 1997. prostate-specific antigen in predicting prostate histology. -_ 11. Richardson, T. D. and Oesterling, J. E.: Age-specific reference Urology, suppl., 48:23,1996. ranges for serum prostate-specific antigen. Urol. Clin. N. 29. Prestigiacomo, A. F., Lilja, H., Pettersson, K., Wolfert, R. L. and h e r . , 24: 339, 1997. Stamey, T. A.: A comparison of the free fraction of serum 12. Kalish, J., Cooner, W. H. and Graham, S. D., Jr.: Serum PSA prostate specific antigen in men with benign and cancerous adjusted for volume of transition zone (PSAT) is more accurate prostates: the best case scenario. J. Urol., 156 350, 1996. than PSA adjusted for total gland volume (PSAD)in detecting 30. Partin, A. W., Catalona, W. J., Southwick, P. C., Subong, E. N., adenocarcinoma of the prostate. Urology, 43:601,1994. Gasior, G. H. and Chan, D. W.: Analysis of percent free 13. Kurita, Y.,Ushiyama, T., Suzuki, K, Fujita, K. and Kawabe, K.: prostate-specific antigen (PSA) for prostate cancer detection: PSA value adjusted for the transition zone volume in the influence of total PSA, prostate volume, and age. Urology, diagnosis of prostate cancer. Int. J. Urol., 3 367,1996. suppl., 48:55,1996. 14. Zlotta, A. R., Djavan, B., Marberger, M. and Schulman, C. C.: Prostate specific antigen density of the transition zone: a new EDITORIAL COMMENT effective parameter for prostate cancer prediction. J . Urol., 167: 1315,1997. The relationship between serum PSA and prostate epithelial vol15. Hammerer, P. G., Mc Neal, J . E. and Stamey, T. A.: Correlation ume was first shown by Lepor et a1 (reference 16 in article), and led between serum prostate specific antigen levels and the volume to the observation that volume of the transition zone may be predicof the individual glandular zones of the human prostate. tive of serum PSA levels (references 14 and 15 in article). This J . Urol., 153. 111, 1995. 16. Lepor, H., Wang, B. and Shapiro, E.: Relationship between pros- predictive ability is clearly a function of total prostatic volume, with tatic epithelial volume and serum prostate-specific antigen larger prostates usually having a greater epithelial volume and, therefore, a greater contribution to serum PSA. The authors of this levels. Urology, 61: 199,1994. 17. Morgan, T.O.,Jacobsen, S. J., McCarthy, W. F., Jacobson, D. J., report first observed that use of PSA density calculated using the McLeod, D. G. and Moul, J. W.: Age-specific reference ranges volume of the transition zone rather than the total prostate provided for prostate-s&c antigen in black men. New Engl. J . Mid., greater specificity in determining the likelihood of prostate cancer than serum PSA alone (reference 14 in article). They compared the 335:304, 1996. 18. Creasy, T., Lesna, M., Rundle, D., Bramble, J . and Morley, R.: use of PSA transition zone to more conventional parameters, such as Central gland PSA density-a more sensitive tool for the di- PSA density, PSA velocity and free-to-total PSA ratio. Each of these agnosis of carcinoma of the prostate. J. Urol., part 2,167: 55, parameters has been shown to reduce the need for unnecessary biopsies in patients with elevated serum PSA but often a t the cost of abstract 209,1997. 19. Partin, A. W. and Oesterling, J . E.: The clinical usefulness of a reduced sensitivity for cancer detection.’ The free-to-total ratio has prostate specific antigen: update 1994. J. Urol., 152 1358, received the most recent attention. In addition to reducing the need for unnecessary biopsies, many experts use this parameter to iden1994. 20. Horninger, W.,Reissigl, A., Klockler, H., Holtl, L., Fink, K. and tify a high risk group that requires additional biopsies. The authors designed a large prospective multicenter trial which Bartsch, G.: Improvement of early detection of prostate cancer by using PSA-transitional zone density (PSA-TZ density) and critically examines the ability of various PSA parameters to remove percent free PSA in addition to total PSA levels. J . Urol., part the need for unnecessary biopsies. Of 559 patients undergoing biopsy 217 had prostate cancer on completion of 2 sets of biopsies. Using a 2,167: 116,abstract 454, 1997. 21. Maeda, H.,Ishitoya, S., Maekawa, Y., Aoki, Y., Okubo, K., 95% sensitivity for prostate cancer detection 47% of unnecessary Okada, T., Suzuki, Y. and Arai, Y.: Prostate specific antigen biopsies would have been avoided if using PSA transition zone comdensity of the transition zone in the detection of prostate pared to 32% free-to-total PSA. The authors have provided sensitivcancer. J. Urol., part 2,167: 58,abstract 219, 1997. ity analyses for each of the parameters to allow the clinician to 22. Gohji, K,Nomi, M., Egawa, S., Morisue, K , Takenaka, A,, determine clinically useful cutoff levels. At the suggested cutoff of Okamoto, M., Ohori, M. and Fuj, A.: Detection of prostate 0.25 ngJml.lcc, PSA transition zone carries a positive predictive carcinoma using prostate specific antigen, its density, and the value of 53% and a n impressive negative predictive value of 94%. The density of the transition zone in Japanese men with interme- negative predictive value may ultimately be most important to diate serum prostate specific antigen concentrations. Cancer, the clinician in determining if a biopsy can be avoided. Also impres7% 1969,1970. sive is the fact that the predictive power of PSA transition zone is not 23. Catalona, W.J., Richie, J. P., deKernion, J. B., Ahmann,F. R., affected by age or the presence of abnormalities on digital rectal RatlifT, T. L., Dalkin, B.L., Kavoussi, L. R., MacFarlane, M. T. examination. 5. Morote, J., Raventajs, C. X., Lorente, J. A., Lopez-Pacios, M. A.,

~~

~

PROSTATE SPECIFIC ANTIGEN DENSITY OF TRANSITION ZONE Of some concern may be the fact that the predictive power of PSA transition zone is significantly affected by prostate size. In prostates less than 30 gm. PSA transition zone is less useful a t the suggested cutoff than free-to-total PSA, which may have ultimately affected the overall PSA. Transition zone predictive values in this study since 76% of patients biopsied had prostates larger than 30 g m . This finding may not be consistent with the population we biopsy in the United States, particularly in view of the trend towards detecting prostate cancer in younger patients. Finally, unlike other PSA parameters, such as free-to-total PSA, PSA transition zone requires the use of transrectal ultrasound, escalating the cost of the test a t many centers. Many may believe that if the patient is to undergo transrectal ultrasound, there is minimal added morbidity to performing the biopsy to exclude cancer and eliminate the need for future transrectal ultrasound procedures. This belief is particularly true in view

419

of recent evidence that 2 sets of sextant cores (12 cores) can be performed at 1 sitting without increasing morbidity.* Ultimately, the deciding factor at individual centers may be the saved cost of eliminating the need for biopsy. The cost savings per undiagnosed cancer may ultimately not be justified in younger men. Samir S. Taneja and Herbert Lepor New York University Medical Center New York,New York

1. Taneja, S. S.: Editorial: refining the gold standard-can we improve on serum PSA? J. Urol., 157: 1752, 1997. 2. Levine, M. A., Melamed, J., Ittmann, M. and Lepor, H.: Two consecutive sets of transrectal ultrasound guided sextant biopsies of the prostate for the diagnosis of prostate cancer. J. Urol., 159 471,1998.