androgen receptor, ki67, and p53 expression in radical prostatectomy specimens predict treatment...
TRANSCRIPT
OmbMbctaRHcspsCl3
SgAJt
TE
SoCJpU
Uho
r
ADULT UROLOGY
3
ANDROGEN RECEPTOR, Ki67, AND p53 EXPRESSIONIN RADICAL PROSTATECTOMY SPECIMENS PREDICTTREATMENT FAILURE IN JAPANESE POPULATION
TAKAHIRO INOUE, TAKEHIKO SEGAWA, TAIZOU SHIRAISHI, TORU YOSHIDA,YOSHINOBU TODA, TOMOMI YAMADA, NAOKO KINUKAWA, HIDEFUMI KINOSHITA,
TOSHIYUKI KAMOTO, AND OSAMU OGAWA
ABSTRACTbjectives. To evaluate multiple known prognostic markers in localized prostate cancer using tissueicroarrays in Japanese patients. Molecular studies have suggested that ethnicity influences prostate tumoriology.ethods. Specimens were studied from 52 patients who underwent radical surgery at our institution
etween 1997 and 2001 without neoadjuvant hormonal therapy and with three or more available andomplete cancer spots. Ki67, p53, and androgen receptor antigen expression were examined. Immunohis-ochemical scores were compared with outcomes of chemical relapse as monitored using prostate-specificntigen.esults. Pathologic tumor classification (P � 0.047), World Health Organization score (P � 0.026), Worldealth Organization histologic grade (P � 0.026), and surgical margin status (P � 0.018) were significantonventional clinicopathologic variables for predicting biochemical failure. The tissue microarray Gleasonum (P � 0.038), tissue microarray primary Gleason grade (P � 0.013), Ki67 labeling index (P �0.0001),53 (P � 0.0097), and androgen receptor (P � 0.0113) antigen expression also were significant. Moreover,urgical margin status and Ki67 labeling index were independently associated with treatment failure.onclusions. Especially together, the Ki67 labeling index and p53 and androgen receptor expression in
ocalized prostate cancer often predicted postoperative progression in Japanese patients. UROLOGY 66:32–337, 2005. © 2005 Elsevier Inc.
oucsoHTtctmbtttJptf
ubstantial differences exist in the incidence ofclinical prostate cancer (PCa) among ethnic
roups, with black Americans having a 10-fold andmerican whites a 5-fold greater incidence than
apanese men.1,2 Despite the striking racial varia-ion in the incidence of clinical PCa, the prevalence
his work was supported by a Grant-in-Aid from the Ministry ofducation, Science, Sports and Culture of Japan.From the Department of Urology, Kyoto University Graduate
chool of Medicine, Kyoto, Japan; Department of Second Pathol-gy, Mie University School of Medicine, Tsu, Japan; Anatomicalenter for Kyoto University Graduate School of Medicine, Kyoto,
apan; Division of Medical Informatics, Kyusyu University Hos-ital, Kyusyu, Japan; Department of Urology, Kansai Medicalniversity, Moriguchi, JapanReprint requests: Osamu Ogawa, M.D., Ph.D., Department of
rology, Kyoto University Graduate School of Medicine, 54 Kawa-aracho, Shogoin, Sakyoku, Kyoto 606-8507, Japan. E-mail:[email protected]
Submitted: November 2, 2004, accepted (with revisions): Feb-
cuary 24, 2005© 2005 ELSEVIER INC.32 ALL RIGHTS RESERVED
f latent carcinoma seems to be similar across pop-lations, suggesting that promotion of microscopicarcinoma to clinically significant cancer differsubstantially among these ethnic groups. More-ver, PCa risk increases for Japanese migrants toawaii and Japanese migrants to Los Angeles.1,2
his epidemiologic evidence emphasizes thathe incidence and progression of PCa is geneti-ally and environmentally influenced. In fact, la-ent tumors in men from Japan showed more rasutations than those from U.S. whites and
lacks.3 Additionally, a difference was noted be-ween Japanese and Americans in the p53 muta-ional spectrum.4 All these facts allow us to assesshe risk of disease progression more accurately inapanese patients with PCa to determine the appro-riate treatment options better. We constructed aissue microarray (TMA) of prostate carcinomarom Japanese patients with comprehensive clini-
al data and evaluated the significance of known0090-4295/05/$30.00doi:10.1016/j.urology.2005.02.028
pa
S
ssta(TatttmhuG2tipcfb(sfoii
T
swcaio
I
mMNwaoaeaAeacmicnieimt
3embtecttc
S
mMtcame
F
AP
FT
P
W
P
W
W
KDu
U
rognostic markers such as Ki67,5,6 p53,7–9 andndrogen receptor (AR).10
MATERIAL AND METHODS
PECIMENS AND CLINICOPATHOLOGIC FEATURESThe entry criteria for this retrospective cohort study to con-
truct a PCa tissue array included clinically localized PCa re-ected between June 1997 and August 2001, whole mountissue specimens processed by standard histologic methodsnd available at our institution, and PCa of sufficient sizegreater than 5 mm) present in the specimens to be cored forMAs. A total of 80 patients with PCa fulfilled these criteria,nd their specimens were cored to produce TMAs. Amonghese 80 patients, 52 were enrolled in this study and had methe following criteria: no preoperative treatment, and at leasthree available cancer spots for evaluation of all moleculararkers. All prostate specimens were processed by standard
istologic methods.11 All the specimens were graded by two ofs (Taizou S. and Takahiro I.) independently according to theleason method and staged pathologically according to the002 American Joint Committee on Cancer TNM classifica-ion and the Japanese General Rules for Clinical and Patholog-cal Studies on Prostate cancer (third edition). All clinical andathologic data were obtained from the medical records. Thelinicopathologic data are summarized in Table I. The date ofailure was considered the time of the initial postoperativelood sampling yielding detectable prostate-specific antigenPSA; 0.1 mg/mL or greater). When the serum PSA level afterurgery did not decline to less than 0.1 ng/mL, the date ofailure was defined as the time of surgery. The most recent datef follow-up was December 31, 2003. All the patients involvedn this study provided informed consent before participatingn this investigation.
MA CONSTRUCTIONAll samples were arrayed, as previously described, with
ome modifications.12–15 Tissue cylinders 1.0 mm in diameterere punched from the selected areas of each block by a pre-
ision instrument (Beecher Instruments, Silver Spring, Md)nd were embedded into paraffin blocks in a systematic fash-on. For each case, three cores with the most undifferentiatedr the main tumor area were selected and arrayed.
MMUNOHISTOCHEMISTRYStandard indirect immunoperoxidase procedures usingonoclonal antibodies were applied to detect Ki67 (1:100,IB-1, DAKO, Kyoto, Japan), p53 (1:400, Do7, Novocastra,ew Castle, UK), and AR (1:100, 2F12 Novocastra). Tumorsith known positivity were used as positive controls for all
ntibodies. As negative controls, the primary antibodies weremitted. All slides were evaluated by two of us (Takehiko S.nd Takahiro I.) independently. Nuclear staining was consid-red representative for Ki67 and p53, and both cytoplasmicnd nuclear staining of epithelium was considered positive forR. Therefore, we defined the Ki67 labeling index and p53xpression as the percentage of nuclear area stained with thesentibodies. One to three nonoverlapping measurements, in-luding the most intense staining area, were made at highagnification (�200) in each cancer spot. Each measurement
ncluded at least 100 epithelial cell nuclei, which resulted inounts including 300 to 1200 nuclei for each case. AR immu-opositivity was graded semiquantitatively as weak (no stain-
ng or light immunostaining involving less than 10% of thepithelium), moderate (light-to-moderate immunostainingnvolving 10% to 50%), and strong (moderate-to-strong im-
unostaining involving more than 50%). Thereafter, we de-
ermined a score as follows: strong AR staining was scored as cROLOGY 66 (2), 2005
; moderate AR staining as 2; and weak AR staining as 1. Tovaluate the prognostic value of each marker subjected to im-unostaining, the mean Ki67 labeling index (sum of the la-
eling index of each cancer spot in each case divided by theotal number of evaluated spots), maximal percentage of p53xpression (greatest percentage of expression among all can-er spots in each case), and mean AR staining score (sum ofhe AR score of each cancer spot in each case divided by theotal number of evaluated spots) were determined and re-orded for each case.
TATISTICAL ANALYSISThe survival time from the date of prostatectomy to treat-ent failure or last follow-up was estimated using the Kaplan-eier method. A log-rank test was used to examine the rela-
ionship between each molecular marker, histologic andlinical data, and PSA relapse-free survival. In multivariatenalysis with a Cox proportional hazards model, a stepwiseethod was used to determine the parameters with the great-
st influence on the risk of progression. Spearman’s rank order
TABLE I. Demographic and clinicopathologicfeatures of 52 patients
eature
ge (yr) 50–76 (68; 64, 72)reoperative PSA
(ng/mL)3.8–120 (9.95; 7.75, 17.15)
ollow-up (days) 727–2343 (1274; 1039, 1759)ime to PSA failure
(mo)0–648 (157; 0, 415)
athologic T stage (n)pT2 14pT3 38hole mount tissue
Gleason sum (n)5 26 27 458 09 310 0
rimary Gleason score3 404 115 1HO score2 13 194 275 26 27 1HO histologic grade0 11 192 273 34 2
EY: PSA � prostate-specific antigen; WHO � World Health Organization.ata presented as ranges, with median and 25th and 75th percentiles in parentheses,nless otherwise noted.
orrelation analysis was used to analyze the statistical signifi-
333
ceasw
C
wTy
HM
g(
o4Ttps1fapdguK0a(
M
ptlwP
Pialdasw5speb
swmmrcmfst
3
ance of the correlations among the Ki67 labeling index, ARxpression, and p53 expression in cancer spots. Statisticalnalysis was performed using StatView, version 5.0 (SAS In-titute, Cary, NC). All P values were two-tailed, and P �0.05as considered to indicate significance.
RESULTS
LINICAL AND HISTOPATHOLOGIC PARAMETERS
The known clinical and pathologic variables ofhole mount prostatic tissues were dichotomized.able II summarizes the results of univariate anal-sis for disease-free survival.
ISTOPATHOLOGIC PARAMETERS AND MOLECULAR
ARKERS IN TMABoth the TMA Gleason sum and TMA Gleason
rade significantly influenced PSA-free survivalP � 0.038 and P � 0.0013, respectively).In the 52 cases analyzed in this study, the range
f the Ki67 labeling index was 0 to 34.3 (median.475, 25th percentile 2.38, 75th percentile 7.77).he percentage of p53 positive expression was 0%
o 90% (median 11%, 25th percentile 2.165%, 75thercentile 23.67%). The range of the AR expres-ion score was 1 to 3 (median 1.5, 25th percentile.0, 75th percentile 1.835). The labeling indexesor Ki67 and p53 expression were dichotomizedccording to a cutoff of 5 and 15, respectively, asreviously reported.5–9 AR immunostaining wasichotomized as high (mean AR score 1.6 orreater) vs. low (mean AR score less than 1.6). Bynivariate dichotomized log-rank analysis, thei67 labeling index (P �0.0001), p53 (P �.0097), and AR (P � 0.0113) were significantlyssociated with PSA-defined disease-free survival
TABLE II. Univariate anatreatm
Factor Investigated
Age (�70 vs. �70 yr)Preoperative PSA (�10 vs. �10 ng/mL)pT stage (�T3 vs. �T3)Gleason sum (�7 vs. �7)Primary Gleason score (�4 vs. �4)WHO score (�4 vs. �4)WHO histologic grade (�2 vs. �2)Surgical margin (positive vs. negative)Perineural invasion (positive vs. negativeTMA Gleason sum (�7 vs. �7)TMA primary Gleason grade (�4 vs. �4p53 expression (positive vs. negative)Ki67 LI (positive vs. negative)AR (high vs. low)
KEY: PSA � prostate-specific antigen; WHO � World HealAR � androgen receptor.
Table II and Fig. 1). a
34
ULTIVARIATE ANALYSIS OF ALL VARIABLES
In stepwise multivariate analysis with a Cox pro-ortional hazards model, the surgical margin sta-us (P � 0.0263, hazards ratio 5.299) and Ki67abeling index (P � 0.0004, hazards ratio 7.517)ere independent, significant determinants ofSA-defined disease-free survival.
COMMENT
This report is of an initial study of the utility ofCa TMAs in a Japanese population. Our study
ncluded a small number of patients. Nevertheless,ccording to the assumption that at a significanceevel of less than 5% and a power of 80%, we canetect a statistically significant difference betweengood prognostic group with a 3-year PSA-free
urvival rate of 80% and a poor prognostic groupith a 3-year PSA-free survival rate of 40% using4 patients (27 patients for each group), we had aufficient number of patients in our study at 52atients. Hence, our sample number permitted ad-quate discrimination of the statistical significanceetween our subsets of analysis.According to univariate analysis, the Gleason
um was not related to PSA failure, inconsistentith the findings of many other reports.16,17 Thisight have been because the histologic results ofost of our patients were Gleason sum 7 or less,
esulting in an imbalance of numbers in each di-hotomized category. Considering that the pri-ary Gleason score was marginally related to PSA
ailure, if more patients were analyzed in futuretudies, the same results as in previous reports inerms of the relationship between the Gleason sum
is of time to PSA-definedfailure
No. of CasesNo. of PSA Failure Cases) P Value
34 (12) vs. 18 (8) 0.66526 (8) vs. 26 (12) 0.28614 (2) vs. 38 (18) 0.0474 (1) vs. 48 (19) 0.500
40 (13) vs. 12 (7) 0.06120 (4) vs. 32 (16) 0.02620 (4) vs. 32 (16) 0.026
36 (18) vs. 16 (2) 0.01829 (13) vs. 23 (7) 0.19617 (3) vs. 35 (17) 0.03825 (4) vs. 27 (16) 0.001
23 (13) vs. 29 (7) 0.009724 (16) vs. 28 (4) �0.000126 (14) vs. 26 (6) 0.0113
anization; TMA � tissue microarray; LI � labeling index;
lysent
(
)
)
th Org
nd PSA failure might be seen.
UROLOGY 66 (2), 2005
swranpccmp
msupTosfcgipt
ifita
ntwhsippstptds(phdS(tcwddpdpicstp
dopd
Fcp(
U
On multivariate analysis, our data showed thaturgical margin status and Ki67 labeling indexere independent predictors of PSA failure after
adical prostatectomy. Several other reports havelso revealed that positive surgical margins are sig-ificantly associated with the risk of biochemicalrogression after radical prostatectomy.18,19 Re-ently, Shuford et al.20 reported that patients withapsular incision after radical prostatectomy wereore likely to have biochemical recurrence than
IGURE 1. Kaplan-Meier PSA-free survival curve ac-ording to (A) Ki67 labeling index (LI) (P �0.0001), (B)53 expression (P � 0.0097), and (C) AR expressionP � 0.0113).
atients with pT2 and pT3 and negative surgical t
ROLOGY 66 (2), 2005
argins, even after adjusting for the effect of Glea-on score, preoperative PSA level, and tumor vol-me. This finding shows that residual cancer at therostatic bed is an adverse prognostic indicator.herefore, there is no doubt that among the vari-us conventional clinicopathologic factors, onlyurgical margin status remains as an independentactor after multivariate analysis. Positivity in-luded both incision into the capsule by the sur-eon and tumors extending to the edge of the spec-men without adequate periprostatic tissue torovide a histologic diagnosis of extraprostatic ex-ension.The usefulness of the Ki67 labeling index as an
ndependent prognostic marker has been con-rmed in previous reports,5,6,21 making this pro-ein clinically applicable to PCa in Japanese, as wells other populations.Our results showed p53 to be an important prog-
ostic marker for patients with localized PCareated by radical prostatectomy, in agreementith recent studies finding p53 protein immuno-istochemical overexpression to have prognosticignificance in PCa.7–9 We used spots with a max-mal percentage of p53 staining as representative of53 staining for each case, because p53 overex-ression is an aggressive feature of cancer.7–9 Con-idering the mean percentage of p53 as representa-ive staining, a relationship was found between therognosis and the mean value, although this rela-ionship was not statistically significant (P � 0.11,ata not shown). In the present study, p53 nucleartaining was demonstrated in only 3 patients21.4%) with less than Stage pT3 disease (StageT2 or less). Moreover, none of these 3 patientsad PSA failure. Conversely, 20 patients (52.6%)emonstrated a positive p53 status with a tumor oftage pT3 or greater. Among them, 13 patients65%) had PSA failure. Therefore, p53 accumula-ion is associated with locally high advanced stageancer. Moreover, considering that a reduction ofild-type p53 expression in an androgen-depen-ent cell line, LNCaP, induced androgen-indepen-ent proliferation, some PCa cells with nuclear53 accumulation might pose androgen indepen-ence.22 Although conflicting data have been re-orted regarding whether p53 status is prognosticn PCa, especially in the United States,23 and someonflicts exist concerning analyzing p53 expres-ion using TMAs,24 we demonstrated the prognos-ic value of p53 staining in our TMAs of a Japaneseopulation.High AR protein expression predicted shorter
isease-free survival in our investigation. More-ver, Ayala et al.25 recently reported high AR ex-ression in PCa to be associated with aggressiveisease. Using the highest AR expression score as
he representative of each case, we considered the335
saTptt(apteiwpegcaafitts
rpvisiptpvsegal
fFcrsfdJ
ec
o2
oct
mtC
ic4
im9
edc
pt1
nv6
sac
dd4
ms
rh
rp
cy
aps
(rJ
tr
tn
pp
pa
3
tatistical analysis to be under-power for the imbal-nce of numbers of each category dichotomized.herefore, in this study, we used the mean AR ex-ression score as representative of each case, al-hough a significant correlation was also found be-ween the highest AR score and the prognosisP � 0.0464, data not shown). Although the mech-nisms underlying poor prognosis in AR overex-ression are not fully understood, considering thathe androgen/AR complex may regulate cell prolif-ration and survival of the prostate epithelium,26 its conceivable that high AR expression correlatesith aggressive disease. Recently, Chen et al.27 re-orted that AR overexpression induced cell prolif-ration at low androgen concentrations in andro-en-dependent cell lines, which suggests that PCaells with AR overexpression might be highlydaptable to a low androgen environment due toging.28 Moreover, this aggressiveness was con-rmed by a significant positive relationship be-ween AR immunostaining and p53 expression orhe Ki67 labeling index in each of our TMA cancerpots (data not shown).In this study, 1 (10%) of 10 patients developed
ecurrence with a negative Ki67 labeling index andositive p53 expression. Moreover, no patient de-eloped recurrence with a negative Ki67 labelingndex and positive AR expression. This might giveome explanation of why only the Ki67 labelingndex of the three molecular markers was an inde-endent factor on multivariate analysis. Our mul-ivariate analysis showed that Ki67 labeling indexositivity was a consequence of earlier events in-olving other protein expression abnormalitiesuch as p53 or AR overexpression. Nevertheless,xamination of the other molecular markers to-ether with the Ki67 labeling index might be desir-ble to delineate an intermediate risk group withonger follow-up.
CONCLUSIONS
We have demonstrated that our PCa TMA is use-ul for evaluating survival-related tissue markers.urthermore, Ki67, p53, and AR expression in lo-alized PCa is useful for predicting recurrence afteradical prostatectomy. Admittedly the presenttudy included fewer patients than many papersrom Western countries, indicating a need for ad-itional information with more markers in a largeapanese study population.
REFERENCES1. Taylor JD, Holmes TM, and Swanson GM: Descriptive
pidemiology of prostate cancer in metropolitan Detroit. Can-er 73: 1704–1707, 1994.
2. Oishi K, Yoshida O, and Schroeder FH: The geographyf prostate cancer and its treatment in Japan. Cancer Surv 23:
67–280, 1995. c36
3. Shiraishi T, Muneyuki T, Fukutome K, et al: Mutationsf ras genes are relatively frequent in Japanese prostate can-ers: pointing to genetic differences between populations. An-icancer Res 18: 2789–2792, 1998.
4. Watanabe M, Ushijima T, Kakiuchi H, et al: p53 geneutations in human prostate cancers in Japan: different mu-
ation spectra between Japan and Western countries. Jpn Jancer Res 85: 904–910, 1994.
5. Bubendorf L, Sauter G, Moch H, et al: Ki67 labelingndex: an independent predictor of progression in prostateancer treated by radical prostatectomy. J Pathol 178: 437–41, 1996.
6. Bubendorf L, Tapia C, Gasser TC, et al: Ki67 labelingndex in core needle biopsies independently predicts tu-
or-specific survival in prostate cancer. Hum Pathol 29:49 –954, 1998.
7. Bauer JJ, Sesterhenn IA, Mostofi FK, et al: Elevated lev-ls of apoptosis regulator proteins p53 and bcl-2 are indepen-ent prognostic biomarkers in surgically treated clinically lo-alized prostate cancer. J Urol 156: 1511–1516, 1996.
8. Leibovich BC, Cheng L, Weaver AL, et al: Outcomerediction with p53 immunostaining after radical prostatec-omy in patients with locally advanced prostate cancer. J Urol63: 1756–1760, 2000.
9. Kuczyk MA, Serth J, Bokemeyer C, et al: The prog-ostic value of p53 for long-term and recurrence-free sur-ival following radical prostatectomy. Eur J Cancer 34:79 – 686, 1998.10. Henshall SM, Quinn DI, Lee CS, et al: Altered expres-
ion of androgen receptor in the malignant epithelium anddjacent stroma is associated with early relapse in prostateancer. Cancer Res 61: 423–427, 2001.
11. Inoue T, Hioki T, Hayashi N, et al: Preoperative pre-ictors of cancerous involvement of the neurovascular bun-les in patients with localized prostate cancer. Int J Urol 9:7–53, 2002.12. Kononen J, Bubendorf L, Kallioniemi A, et al: Tissueicroarrays for high-throughput molecular profiling of tumor
pecimens. Nat Med 4: 844–847, 1998.13. Bubendorf L, Nocito A, Moch H, et al: Tissue microar-
ay (TMA) technology: miniaturized pathology archives forigh-throughput in situ studies. J Pathol 195: 72–79, 2001.14. Torhorst J, Bucher C, Kononen J, et al: Tissue microar-
ays for rapid linking of molecular changes to clinical end-oints. Am J Pathol 159: 2249–2256, 2001.15. Rubin MA, Dunn R, Strawderman M, et al: Tissue mi-
roarray sampling strategy for prostate cancer biomarker anal-sis. Am J Surg Pathol 26: 312–319, 2002.
16. Roehl AK, Han M, Ramos GC, et al: Cancer progressionnd survival rates following anatomical radical retropubicrostatectomy in 3478 consecutive patients: long-term re-ults. J Urol 172: 910–914, 2004.
17. Han M, Partin WA, Zahurak M, et al: Biochemicalprostate specific antigen) recurrence probability followingadical prostatectomy for clinically localized prostate cancer.Urol 169: 517–523, 2003.18. van den Ouden D, Bentvelsen FM, Boeve ER, et al: Posi-
ive margins after radical prostatectomy: correlation with localecurrence and distant progression. Br J Urol 72: 489–494, 1993.
19. Paulson DF, Moul JW, Walther JP, et al: Radical pros-atectomy for clinical stage T1-2N0M0 prostatic adenocarci-oma: long-term results. J Urol 144: 1180–1184, 1990.20. Shuford DM, Cookson SM, Chang SS, et al: Adverse
rognostic significance of capsular incision with radical retro-ubic prostatectomy. J Urol 172: 119–123, 2004.21. Keshgegian AA, Johnston E, and Cnaan A: Bcl-2 onco-
rotein positivity and high MIB-1 (Ki-67) proliferation ratere independent predictive markers for recurrence in prostate
arcinoma. Am J Clin Pathol 110: 443–449, 1998.UROLOGY 66 (2), 2005
tp2
f3
obR
td
c
m3
s
U
22. Burchardt M, Burchardt T, Shabsigh A, et al: Reduc-ion of wild-type p53 function confers a hormone resistanthenotype on LNCaP prostate cancer cell. Prostate 48:25–230, 2001.23. Brooks DJ, Bova SG, Ewing MC, et al: An uncertain role
or p53 gene alterations in human prostate cancers. J Urol 56:814–3822, 1996.24. Merseburger AS, Kuczyk MA, Serth J, et al: Limitations
f tissue microarrays in the evaluation of focal alterations ofcl-2 and p53 in whole mount derived prostate tissues. Oncol
ep 10: 223–228, 2003. JROLOGY 66 (2), 2005
25. Ayala G, Li R, Weigel N, et al: High levels of androgen recep-or in PCA is associated with aggressive disease and is an indepen-ent predictor of survival. Am J Surg Pathol 28: 928–934, 2004.26. Isaacs JT: Antagonistic effect of androgen on prostatic
ell death. Prostate 5: 545–557, 1984.27. Chen DC, Welsbie SD, Tran C, et al: Molecular deter-inants of resistance to antiandrogen therapy. Nat Med 10:
3–39, 2003.28. Iwamoto T, Yanase T, Koh E, et al: Reference ranges of
erum total and free testosterone in Japanese male adults. Jpn
Urol 95: 751–760, 2004.337