Abstract

Background: Pathways involved in androgen metabolism have been implicated in the etiology of prostate cancer. The goal of this study was to evaluate the effect of CYP3A4, a gene associated with the oxidative deactivation of testosterone, on the clinical presentation of prostate cancers. Methods: A polymerase chain reaction-based approach was used to identify sequence variants of the human CYP3A4 gene. To ascertain whether allelic variants of the CYP3A4 gene were associated with tumor stage and grade and age of the patient at diagnosis, we determined CYP3A4 genotypes in 230 Caucasian men with incident prostate cancer. Results: We identified a novel genetic variant (CYP3A4-V) that has an altered 5′ regulatory element, containing an A to G mutation, upstream of the CYP3A4 gene. We then compared clinical characteristics of prostate cancers in men who did and did not carry this variant. The presence of the CYP3A4-V allele was associated with a higher tumor-lymph node-metastasis (TNM) stage and Gleason grade. The association between CYP3A4 genotype and tumor stage was most pronounced in men diagnosed at a relatively old age who reported no family history of prostate cancer. In this group, 46% of men with stage T3/T4 tumors carried CYP3A4-V, whereas only 5% of individuals with stage T1 tumors carried CYP3A4-V (adjusted odds ratio = 9.45; 95% confidence interval = 2.54–35.17; x2 1 = 12.28; two-sided P<.001). Conclusions: We determined that a single base change in the 5′ flanking region of the CYP3A4 gene was associated with higher clinical stage and grade in men with prostate tumors. Our results suggest that mutations in the CYP3A4 gene may influence prostate carcinogenesis. [J Natl Cancer Inst 1998;90: 1225-9]

The etiology of prostate cancer involves the effects of androgens as well as of inherited genotypes that may regulate androgen metabolism. Candidate prostate cancer genes include those involved in androgen metabolism, such as the androgen receptor (1,2) or 5α-reductase type II (3) genes. Additional candidates include members of the cytochrome P450 supergene family involved in androgen metabolism. One member of this multigene family is CYP3A4, a gene involved in the oxidation of testosterone to 2β-, 6β-, or 15b-hydroxytestosterone (4). Substantial interindividual variability in metabolism of specific compounds by the Cyp3a4 protein has been reported (5), yet no genetic basis for this variability has been found. It has also been reported that the Cyp3a4 protein is expressed in only 61% of prostate tumors (6), which suggests that there is tumor-specific variability in Cy3a4 protein expression.

The goal of this study was to identify germline genetic variants in the CYP3A4 gene that may be used as molecular biomarkers in studies of prostate cancer susceptibility. We searched for variants in the CYP3A4 gene and identified a previously unreported variant in a regulatory element in the 5′ regulatory region of the CYP3A4 gene. We then examined the relationship between variant CYP3A4 genotypes and the clinical presentation of prostate cancer.

Subjects and Methods

Sample Subjects and Biosample Collection

A reference panel (n 4 94) consisting of healthy, unrelated, Caucasian, male volunteers with no history of cancer at any site was assembled at the Hospital of the University of Pennsylvania, Philadelphia, and was used to identify variants in CYP3A4. The mean age of these men was 63.4 years (standard deviation [SD]412.3 years; range423-89 years). A sample of 230 patients representing incident prostate cancer case subjects was identified through the Urologic Oncology Clinics at the Hospital of the University of Pennsylvania during the period from July 1994 through November 1997. Because of small numbers in other ethnic groups, only non- Hispanic, Caucasian men were included in this study. Men were excluded from this study if they had any history of exposure to finasteride at the time of their prostate cancer diagnosis. Patients who were nonincident case subjects (i.e., those diagnosed >12 months before the date of study ascertainment) were also excluded. The mean age at diagnosis was 63.3 years (SD47.8 years; range445-90 years). Written, informed consent was obtained from all participants in the context of a study protocol (#3614-2) approved by the Committee on Studies Involving Human Beings at the University of Pennsylvania.

Genomic DNA for this study was collected by each study subject using sterile cheek swabs (Cyto- Pak Cytosoft Brush; Medical Packaging Corporation, Camarillo, CA). We processed the genomic DNA by using a protocol modified from that of Richards et al. (7). Briefly, the swab brush was placed inside a 1.5-mL microcentrifuge tube, and 600 µL of 50 mM NaOH was added. The closed tube was vortexed for 5 minutes and then heated at 95 °C for 10 minutes. Finally, 120 µL of 1 M Tris (pH 8.0) was added, after which the brush was removed and discarded.

Information on clinical characteristics of the patients at diagnosis was abstracted from medical records. These characteristics included clinical and pathologic Gleason grade (8), prostate-specific antigen (PSA) levels at diagnosis, and TNM (tumor- lymph node-metastasis) (9) stage. One hundred forty-five (63%) subjects underwent radical prostatectomy for treatment of their disease. In these patients, tumor stage and grade were determined by histopathologic review, which described the capsular status, percent tumor volume, and seminal vesicle involvement. Tumors from the remaining 85 patients were staged by a combination of measurement of serum PSA levels, digital rectal examination, bone scan, and endorectal magnetic resonance imaging. Of the 13 patients diagnosed with T3c stage tumors, eight of these diagnoses were based only on PSA levels, digital rectal examination, magnetic resonance imaging, and bone scan alone, while five patients also underwent seminal vesicle biopsy. Pathologic grading with the use of the Gleason system was undertaken after six to 11 transrectal ultrasound- guided needle biopsies. The Hospital of the University of Pennsylvania investigators reinterpreted the diagnoses of the original biopsy material made at other institutions. For analysis, three TNMstage variables were considered. First, we considered stages T1a-T1c (denoted T1), T2a-T2c (denoted T2), and T3/T4. Second, we considered a binary stage variable of nonpalpable (stages T1a- T1c) and palpable (stages T2a-c, T3a-c, and T4a-b) disease. Finally, we considered a combined disease classification defined simultaneously by stage and grade, in which tumors of stages T1a-c and of a Gleason grade less than 7 were compared with tumors of stages T2-T4 or T1 and of a Gleason grade greater than or equal to 7. Additional risk factor information was obtained from the subjects by selfreport using a questionnaire. These variables included demographic characteristics, medical history, and family history of prostate cancer.

CYP3A4 Genotype Analysis

Variant alleles were detected by polymerase chain reaction (PCR) amplification of a 592-base-pair (bp) fragment upstream from the coding region of the CYP3A4 gene that included a portion of exon 1 (nucleotides -571 to +22). The primers used in this amplification (5′-AAC AGG GGT GGA AAC ACA AT-3′ and 5′-CTT TCC TGC CCT GCA CAG-3′) were generated from the CYP3A4-specific nucleotide sequence published by Hashimoto et al. (10). The PCR reaction mixture consisted of 10 µL of double-distilled H2O, 5 µL of 10x PCR buffer (The Perkin-Elmer Corp., Foster City, CA), 3 µL of 25 mM Mg2+, 1 µL each of 10 mM deoxynucleoside triphosphates, 5 µL each of 5 mM PCR primers, 10 µL of template DNA, 0.8 µL of Taq polymerase (Amplitaq; The Perkin-Elmer Corp.), and 12.2 µL of double-distilled H2O, for a total volume of 50 µL. The temperature profile for the PCR reaction was one cycle each at 94 °C for 5 minutes and 82 °C for 1 minute, followed by 25 cycles of 94 °C for 1 minute, 66 °C for 1 minute with a 0.5 °C per cycle decrease, and 72 °C for 1 minute. This procedure was followed by eight cycles at 94 °C for 1 minute, 50 °C for 1 minute, and 72 °C for 1 minute and a final single 10-minute cycle at 72 °C

Genotypes were visualized by conformationsensitive gel electrophoresis (CSGE) of the PCR product on a 10% nondenaturing polyacrylamide gel by use of the protocol of Ganguly et al. (11), after staining was done with ethidium bromide. For the identification of homozygous wild-type (W/W), homozygous variant (V/V), and heterozygous (W/V) genotypes with the use of CSGE, two samples were loaded onto the polyacrylamide gel for each subject (Fig. 1). In one well, a 6-mL PCR sample was loaded with 6 µL of PCR-generated homozygous variant (V/V) DNA from a known V/V subject (denoted “+V/V” in Fig. 1). In a second well, a 12-mL PCR sample was loaded onto the gel without the addition of V/V DNA (denoted “-V/V” in Fig. 1). In V/V subjects, a single (homoduplex) band was always observed. In W/V subjects, homoduplex and heteroduplex bands were always observed. Thus, the addition of V/V DNA had no effect on the banding pattern for V/V and W/V subjects. In W/W subjects, a single (homoduplex) band was present in those lanes without V/V DNA. The addition of known V/V DNA to the DNA of subjects with W/W genotypes produced both homoduplex and heteroduplex bands.

Statistical Analyses

Nonparametric methods were used to compare proportions in contingency tables using two-sided Fisher's exact tests (FETs) and/or Kruskal-Wallis x2 statistics (for analysis of continuous variables such as PSA level or Gleason grade). Odds ratios were estimated by use of logistic regression models for polytomous or binary outcome stage or grade data. All odds ratio estimates were adjusted for patient's age at diagnosis and method of prostate cancer detection. The detection method was coded as three binary (yes/no) covariates describing referral for diagnosis because of elevated PSA levels, abnormal results from digital rectal examination at routine screening, and/or the existence of prostate cancer symptoms. Stratified analyses were undertaken to compare genotype effects by family history of prostate cancer and age of the patient at diagnosis. Positive family history of prostate cancer was defined as having at least one first- or second-degree relative with prostate cancer. With the use of this definition, 25% of patients had a positive family history. Therefore, an individual in this study defined as having a positive family history of prostate cancer did not necessarily come from a family with a hereditary pattern of prostate cancer. Age at diagnosis was stratified at the median age of diagnosis (63 years) for the study population to distinguish “earlier age at diagnosis” patients (i.e., those diagnosed at or before 63 years of age) from “later age at diagnosis” patients (i.e., those diagnosed after 63 years of age). The Cochran-Mantel-Haenszel chi-squared test (x2 CMH) for nonzero correlation among strata was used to compare contingency tables stratified by family history of prostate cancer and/or age at diagnosis.

Fig. 1

Analysis of the nifedipine-specific element variant at CYP3A4 by conformation-sensitive gel electrophoresis (CSGE). The appearance of homozygous variants (V/V) and heterozygotes (W/V) is identical in the presence (+) and absence (-) of homozygous variant (V/V) DNA. The addition of V/V DNA to homozygous W/W samples induces the generation of heteroduplexes in CSGE analysis and results in the presence of an additional band.

Fig. 1

Analysis of the nifedipine-specific element variant at CYP3A4 by conformation-sensitive gel electrophoresis (CSGE). The appearance of homozygous variants (V/V) and heterozygotes (W/V) is identical in the presence (+) and absence (-) of homozygous variant (V/V) DNA. The addition of V/V DNA to homozygous W/W samples induces the generation of heteroduplexes in CSGE analysis and results in the presence of an additional band.

Results

Variant Genotype Identification

With the use of a reference sample of 94 unselected, unrelated Caucasians with no history of cancer at any site, 15 carriers of variant alleles (12 heterozygotes and three homozygotes) were identified by CSGE analysis of the 5′ regulatory region of the CYP3A4 gene. These subjects were confirmed by direct sequencing to carry an A to G transition mutation that alters the 10-bp (AGGGCAAGAG to AGGGCAGGAG) nifedipine-specific element (NFSE), located -287 to -296 bp from the transcription start site of the CYP3A4 gene (10). Sequence analysis was also undertaken on a randomly selected set of 12 of the 79 individuals inferred to be homozygous wild-type from CSGE. No nucleotide changes relative to the wild-type sequence were detected in these individuals. Thus, we estimated the CYP3A4 variant allele frequency in the U.S. Caucasian population to be 9.6% (18 variant alleles among 188 chromosomal copies), with an observed heterozygosity of 12.8%. The NFSE is a purine-rich element that has homology with the basic transcription element. The NFSE has been previously identified as a CYP3A4-specific element that is bound by nuclear proteins and falls within a region required for CYP3A4 gene transcription in HepG2 cells (a human hepatoma cell line) (10).

Association of CYP3A4 Genotypes With Prostate Cancer

Analyses were undertaken to evaluate the relationship between CYP3A4 genotypes and clinical characteristics of prostate tumors. We found no association between CYP3A4 genotype and the PSA level at diagnosis in analysis of unadjusted PSA (x2 = 0.22; P = .637) or PSA adjusted for age and mode of prostate cancer detection (x21 = 2.14; P = .143). We also found no significant association between genotype and PSA level in any group defined by age at diagnosis or family history of prostate cancer. Genotype was not associated with an earlier age at diagnosis in unadjusted analysis (x21 4 1.12; P = .290) or in analysis adjusted for method of prostate cancer detection (x21 = 0.71; P = .399).

CYP3A4-V genotypes were overrepresented in tumors of higher stage and grade (Tables 1 and 2). As shown in Table 1, CYP3A4-V genotypes were more common in tumors of higher stage (P = .081 [FET]; x21 4 5.12; P = .024), and there were significant differences in this relationship by family history of prostate cancer and age at diagnosis (x2 CMH 4 4.91; df = 1; P = .027). We observed no association between the CYP3A4 genotype and stage in men with a family history of prostate cancer or whose tumor was diagnosed at an earlier age, possibly as a result of small sample sizes. However, there was a significant effect of genotype in patients diagnosed at a later age (P = .0003 [FET]; x21 = 14.81; P<.001), in those without a family history of prostate cancer (P = .049 [FET]; x21 4 6.12; P = .013), or in those diagnosed at a later age and with a negative family history of prostate cancer (P = .0008 [FET]; x21 = 12.28; P<.001). Age- and detection method-adjusted odds ratios from logistic regression approximated the relative risk of having a tumor of an advanced stage (T3/T4) associated with the CYP3A4-V genotype to be 2.10 (95% confidence interval [CI] = 1.09–4.05). The adjusted odds ratio estimates increased to 2.72 (95% CI = 1.24–5.61) for patients without a family history of prostate cancer, to 6.70 (95% CI = 2.54–17.69) for those diagnosed at a later age, and to 9.45 (95% CI = 2.54–35.17) for those without a family history of prostate cancer who were diagnosed at a later age.

By collapsing the data on stage in Table 1, we also evaluated the effect of genotype on nonpalpable (stage T1) and palpable (stages T2-T4) tumors. A marginally significant relationship between the presence of a CYP3A4-V allele and palpable disease remained in patients without a family history of prostate cancer (P value4.060 [FET]; x2144.02; P = .045), and a significant association remained in patients who were diagnosed at a later age (P = .023 [FET]; x21 4 5.29; P = .021). The adjusted odds ratio estimates in these groups were 4.42 (95% CI = 1.17–16.63) for patients without a family history of prostate cancer and 8.34 (95% CI = 1.06–65.5′) for patients diagnosed at a later age. No relationship with genotype was observed in other groups of patients. This result may be explained in part by small sample sizes in some groups. In all analyses, identical inferences were obtained when the data were stratified by use of age cut points other than the median (e.g., at 60 and 65 years of age).

The CYP3A4-V genotype was also overrepresented in patients with tumors of higher Gleason grade who had no family history of prostate cancer and who were diagnosed at a later age. In this group, 13% of patients whose tumors had a Gleason grade of 6 or less carried the CYP3A4-V allele, compared with 24% of patients whose tumors were Gleason grade 7 or greater (x21 4 16.73; P = .010). However, we detected no significant effect of the CYP3A4 genotype on Gleason grade in the total sample, and there was no difference in the genotypespecific mean Gleason grade in any ageor family history-specific group. As shown in Table 2, the CYP3A4 genotype distinguished tumors defined simultaneously by stage and grade in patients diagnosed at a later age (P = .035 [FET]; x21 4 4.64; P = .031) and distinguished such tumors marginally significantly in patients without a family history of prostate cancer (P = .074 [FET]; x214 3.36; P = .067) but did not distinguish tumors from patients in other groups. Again, identical inferences were obtained when analyses were undertaken with the use of age cut points other than the median. We conclude that the CYP3A4 genotype may have a greater effect on tumor stage than on Gleason grade.

Discussion

We identified a novel genetic variant in a 5′ regulatory element of the human CYP3A4 coding region and report that prostate cancer patients who carry this variant allele have tumors of a higher clinical stage than patients who do not carry this variant. The hypothesis that the Cyp3a4 protein may be involved in modifying the clinical presentation of prostate cancers is supported by knowledge about testosterone metabolism. The Cyp3a4 protein is responsible for 2β-, 6β-, and 15b-hydroxylation of testosterone, which may result in the hormone's functional deactivation (4,12). The CYP3A4 allelic variant reported here is an alteration in a transcriptional regulatory element that may be required for CYP3A4 gene expression (10). We therefore hypothesize that men with CYP3A4-V genotypes may have decreased Cyp3a4 protein activity and, thus, decreased 2b-, 6b-, and 15btestosterone oxidation. This decreased oxidation may in turn increase the bioavailability of testosterone for conversion to its intracellular mediator, dihydrotestosterone, the principal androgenic hormone involved in the regulation of prostate cell growth and function (13). Therefore, it is biologically plausible that the CYP3A4 allelic variant reported here may influence androgen-mediated prostate carcinogenesis and, thus, the presentation of prostate cancers.

We report that the effect of the CYP3A4 genotype on clinical presentation of prostate cancer is more pronounced in men diagnosed at an older age. In aging men, free testosterone levels decline moderately (14), and there is a possible shift in the distribution of testosterone metabolites, e.g., a decrease in 5a over 5b metabolites (15). An explanation for this observation is that CYP3A4-V genotypes may be associated with increased testosterone bioavailability, which may be relatively more important in men who have lower basal testosterone levels (e.g., older men) than in men who have higher basal testosterone levels (e.g., younger men). In other words, having a CYP3A4-V genotype may be associated with relatively higher basal testosterone levels in older men, resulting in prostate cancers of higher stage than those occurring in older men who do not carry the variant allele.

There are two primary analytical limitations in this study. First, the sample size may have been too small to detect significant effects in some groups, particularly those with a family history of prostate cancer. Therefore, strong inferences cannot be made about the effect of the CYP3A4 genotype in individuals with a family history of prostate cancer. Second, numerous statistical tests were performed in the generation of the present results. In this hypothesis-generating study, we used a P value of .05 for all statistical inferences. However, even after applying a more conservative significance level of .003 (i.e., .05/18, where 18 is the number of contingency table tests performed), we were still able to conclude statistical significance in the relationship between the CYP3A4 genotype and tumor stage in patients without a family history of prostate cancer and/or in those with late onset cancers (Table 1).

The identification of the CYP3A4 genotype as a biomarker associated with prostate cancer has potential implications for treatment and prevention of prostate cancer. First, knowledge about the CYP3A4 genotype may provide useful information about prostate cancer treatment or prognosis. It has been reported that the CYP3A4 enzyme is detectable in only 61% of prostate tumors (6), suggesting that there is tumor-specific variability in CYP3A4 gene expression. Therefore, response to hormone therapy may be in part determined by the CYP3A4 genotype or the Cyp3a4 phenotype. Our finding that stage at diagnosis is associated with genotype also implies that knowledge of the CYP3A4 genotype may be of value in evaluating prognosis, since prostate tumor stage is an important predictor of mortality from prostate cancer (16). Second, primary prostate cancer prevention strategies may be enhanced by knowledge about the CYP3A4 genotype. In addition to its effects on testosterone metabolism, the Cyp3a4 protein also oxidizes finasteride (17), an inhibitor of the 5a-reductase involved in the formation of dihydrotestosterone. Since individuals who carry the CYP3A4-V allele may have increased activity in the testosterone-dihydrotestosterone pathway, the CYP3A4 genotype may influence an individual's response to prostate cancer chemoprevention by finasteride.

This case-case study design does not allow us to directly address the role of the CYP3A4 alleles in the etiology of prostate cancer. We are currently undertaking formal case-control studies of the CYP3A4 genotype for that purpose. Other genotypes have been identified previously that may be involved in the etiology of prostate cancer. For example, the HPC1 gene may explain a proportion of hereditary prostate cancers (18). Other genes involved in androgen metabolism have also been implicated in the etiology of prostate cancer. These include the androgen receptor, 5a-reductase type II, and, possibly the 3α-, 3β-, and 17β-hydroxysteroid dehydrogenases (13,19). In combination with our results, these studies suggest that multiple genes involved in androgen metabolism pathways may play a role in prostate cancer etiology.

References

1
Giovannucci
E
Stampfer
MJ
Krithivas
K
Brown
M
Dahl
D
Brufsky
A
, et al.  . 
The CAG repeat within the androgen receptor gene and its relationship to prostate cancer [published erratum appears in Proc Natl Acad Sci U S A 1997;94:8272]
Proc Natl Acad Sci U S A
 , 
1997
, vol. 
94
 (pg. 
3320
-
3
)
2
Ingles
SA
Ross
RK
Yu
MC
Irvine
RA
La
Pera G
Haile
RW
, et al.  . 
Association of prostate cancer risk with genetic polymorphisms in vitamin D receptor and androgen receptor
J Natl Cancer Inst
 , 
1997
, vol. 
89
 (pg. 
166
-
70
)
3
Reichardt
JK
Makridakis
N
Henderson
BE
Yu
MC
Pike
MC
Ross
RK
Genetic variability of the human SRD5A2 gene: implications for prostate cancer risk
Cancer Res
 , 
1995
, vol. 
55
 (pg. 
3973
-
5
)
4
Waxman
DJ
Attisano
C
Guengerich
FP
Lapenson
DP
Human liver microsomal steroid metabolism: identification of the major microsomal steroid hormone 6b-hydroxylase cytochrome P-450 enzyme
Arch Biochem Biophys
 , 
1988
, vol. 
263
 (pg. 
424
-
36
)
5
Kleinbloesem
CH
van
Brummelen P
Faber
H
Danhof
M
Vermeulen
NP
Breimer
DD
Variability in nifedipine pharmacokinetics and dynamics: a new oxidation polymorphism in man
Biochem Pharmacol
 , 
1984
, vol. 
33
 (pg. 
3721
-
4
)
6
Murray
GI
Taylor
VE
McKay
JA
Weaver
RJ
Ewen
SW
Melvin
WT
, et al.  . 
The immunohistochemical localization of drugmetabolizing enzymes in prostate cancer
J Pathol
 , 
1995
, vol. 
177
 (pg. 
147
-
52
)
7
Richards
B
Skoletsky
J
Shuber
AP
Balfour
R
Stern
RC
Dorkin
HL
, et al.  . 
Multiplex PCR amplification from the CFTR gene using DNA prepared from buccal brushes/swabs
Hum Mol Genet
 , 
1993
, vol. 
2
 (pg. 
159
-
63
)
8
Gleason
DF
The Veteran's Administration Cooperative Urologic Research Group: histologic grading and clinical staging of prostatic carcinoma. In: Tannenbaum M, editor
Urologic pathology: the prostate
 , 
1977
Philadelphia
Lea & Febiger
(pg. 
171
-
98
)
9
Beahrs
OH
Henson
DE
Hutter
RV
Kennedy
BJ
Manual for staging of cancer. 4th ed.
 , 
1992
Philadelphia
Lippincott
(pg. 
185
-
6
)
10
Hashimoto
H
Toide
K
Kitamura
R
Fujita
M
Tagawa
S
Itoh
S
, et al.  . 
Gene structure of CYP3A4, an adult-specific form of cytochrome P450 in human livers, and its transcriptional control
Eur J Biochem
 , 
1993
, vol. 
218
 (pg. 
5′5
-
95
)
11
Ganguly
A
Rock
MJ
Prockop
DJ
Conformation- sensitive gel electrophoresis for rapid detection of single-base differences in doublestranded PCR products and DNA fragments: evidence for solvent-induced bends in DNA heteroduplexes [published erratum appears in Proc Natl Acad Sci U S A 1994;91:5217].
Proc Acad Nat Sci U S A
 , 
1993
, vol. 
90
 (pg. 
10325
-
9
)
12
Waxman
DJ
Lapenson
DP
Aoyama
T
Gelboin
HV
Gonzalez
FJ
Korzekwa
K
Steroid hormone hydroxylase specificities of eleven cDNA-expressed human cytochrome P450s
Arch Biochem Biophys
 , 
1991
, vol. 
290
 (pg. 
160
-
6
)
13
Coffey
DS
Androgen action and the sex accessory tissues. In: Knobil E, Neill JJ, Ewing LL, editors
The physiology of reproduction
 , 
1988
New York
Raven Press
(pg. 
1081
-
118
)
14
Vermeulen
A
Androgens in male senescence. In: Nieschlag E, Behre HM, editors
Testosterone: action, deficiency, substitution
 , 
1980
Berlin
Springer-Verlag
(pg. 
261
-
76
)
15
Vermeulen
A
Verdonck
L
Some studies on the biological significance of free testosterone
J Steroid Biochem
 , 
1972
, vol. 
3
 (pg. 
421
-
6
)
16
Johansson
JE
Holmberg
L
Johansson
S
Bergstrom
R
Adami
HO
Fifteen-year survival in prostate cancer. A prospective, population based study [published erratum appears in JAMA[published erratum appears in JAMA 1997;278:206]
JAMA
 , 
1997
, vol. 
277
 (pg. 
467
-
71
)
17
Huskey
SW
Dean
DC
Miller
RR
Rasmusson
GH
Chiu
SH
Identification of human cytochrome P450 isozymes responsible for the in vitro oxidative metabolism of finasteride [published erratum appears in Drug Metab Dispos 1996;24:695].
Drug Metab Dispos
 , 
1997
, vol. 
277
 (pg. 
467
-
71
)
18
Smith
JR
Freije
D
Carpten
JD
Gronberg
H
Xu
J
Isaacs
SD
, et al.  . 
Major susceptibility locus for prostate cancer on chromosome 1 suggested by a genome-wide search
Science
 , 
1996
, vol. 
274
 (pg. 
1371
-
4
)
19
Devgan
SA
Henderson
BE
Yu
MC
Shi
CY
Pike
MC
Ross
RK
, et al.  . 
Genetic variation of 3 beta-hydroxysteroid dehydrogenase type II in three racial/ethnic groups: implications for prostate cancer risk
Prostate
 , 
1997
, vol. 
33
 (pg. 
9
-
12
)

Author notes

Supported by Public Health Service grants ES08031 (National Institute of Environmental Health Sciences) and CA73730 (National Cancer Institute), National Institutes of Health, Department of Health and Human Services (both grants to T. R. Rebbeck); and by the University of Pennsylvania Cancer Center.