Fax: (727) 398-9467
Macrophage migration inhibitory factor evaluation compared with prostate specific antigen as a biomarker in patients with prostate carcinoma
Article first published online: 11 MAR 2002
Copyright © 2002 American Cancer Society
Volume 94, Issue 5, pages 1449–1456, 1 March 2002
How to Cite
Meyer-Siegler, K. L., Bellino, M. A. and Tannenbaum, M. (2002), Macrophage migration inhibitory factor evaluation compared with prostate specific antigen as a biomarker in patients with prostate carcinoma. Cancer, 94: 1449–1456. doi: 10.1002/cncr.10354
- Issue published online: 11 MAR 2002
- Article first published online: 11 MAR 2002
- Manuscript Accepted: 5 NOV 2001
- Manuscript Received: 27 SEP 2001
- Veterans Administration Merit Review
- Bay Pines Veterans Administration Medical Center
- enzyme-linked immunosorbent assay;
- macrophage migration inhibitory factor;
- prostate neoplasms;
- prostate specific antigen
Cytokines are polypeptides that constitute a class of chemical mediator molecules that modulate cell growth by inducing specific target gene expression. The objective of this study was to evaluate the clinical usefulness of serum evaluation of the cytokine macrophage migration inhibitory factor (MIF) in patients undergoing routine prostate specific antigen (PSA) screening.
In this preliminary, retrospective study, the authors report the development of an enzyme-linked immunosorbent assay (ELISA) for MIF determination in serum samples. A polymerase chain reaction (PCR)-based assay investigated associations between MIF expression and prostate carcinoma (CaP). The authors developed a relative quantitative reverse transcriptase-PCR assay to determine MIF mRNA amounts within laser-capture microscopy (LCM)-dissected prostate epithelial cells.
A comparison of serum MIF levels and total PSA levels identified a positive correlation (correlation coefficient [r2] = 0.61; P < 0.001; n = 509 patients), suggesting an association between elevated serum concentrations of these proteins and CaP. A correlation of serum MIF levels with a diagnosis of CaP demonstrated that patients with a previous CaP diagnosis had significantly elevated serum MIF concentrations (mean ± standard deviation, 6.8 ± 0.87 ng/mL; P < 0.001). To associate altered serum MIF levels with MIF mRNA expression within prostate epithelial cells, LCM-dissected prostate epithelial cells (formalin fixed biopsies from three different patients) were used to determine MIF mRNA amounts by PCR analysis. On average, MIF mRNA amounts were 6.5 times higher in CaP epithelial cells that were invasive to the margin compared with MIF mRNA amounts in normal prostate epithelial cells within the same biopsy specimen.
The ELISA data from the current study suggested an association between increased MIF expression and CaP and suggested that serum MIF concentration may serve as a prognostic marker for CaP. Cancer 2002;94:1449–56. © 2002 American Cancer Society.
Prostate carcinoma (CaP) is the most common malignancy in males and is the second leading cause of cancer mortality in the United States. Extreme variation in disease progression is a hallmark of CaP. In some patients, the disease remains localized, whereas, in others, the disease metastasizes quickly. Invasion and metastasis continue to present the greatest obstacles to the successful treatment of patients with CaP. The physician treating patients with this disease has few reliable prognostic markers. In fact, there are few measurable variables that can predict with reliability the clinical behavior of CaP. The current palliative treatment for patients with CaP, which includes hormonal manipulation and/or orchiectomy, often has significant deleterious effects on patient quality of life. These variables make it difficult for the patient and physician to make proper treatment decisions. The first step in potentially rectifying this dilemma is to identify new biomarkers that exhibit a strong association with CaP. To this end, the current study's objective was to determine whether the cytokine macrophage migration inhibitory factor (MIF) is a potential biochemical marker that may be used in management decisions for patients with CaP.
MIF has been described as a pituitary hormone that enters the circulation after infectious or stressful stimuli and acts to counter-regulate glucocorticoid suppression of cytokine production.1 Circulating serum MIF concentrations reportedly increase in patients with disorders like systemic lupus erythematous, glomerulonephritis, uveitis, and rheumatoid arthritis.2–4 Previously, we found that MIF is produced and secreted by in vitro cultured prostate epithelial cells.5 The MIF amounts secreted by in vitro cultured CaP epithelial cell lines are many times greater than the concentrations secreted by in vitro cultured normal prostate epithelial cells.5 Given this information, there is a rationale to suggest that increased serum MIF concentrations would be found in CaP patients. Furthermore, if increased MIF production results in more virulent malignancies, then assessing serum MIF concentration may prove to be a useful prognostic index. Thus, it may be possible to use serum MIF concentration to identify patients who would benefit from aggressive CaP treatment, which ultimately increases the urologic oncologist's ability to apply an appropriate therapy.
Studies originating in this and other laboratories have confirmed that MIF is synthesized and secreted by prostate epithelial cells.5–8 Increasing experimental evidence links elevated MIF expression with carcinoma.9–12 Administration of anti-MIF antibody to tumor-bearing mice significantly suppresses tumor growth and tumor-induced angiogenesis.13–15 The exact physiologic function of MIF in tumor progression is unknown. However, results from our previous studies have correlated increased MIF mRNA expression with the metastatic potential of CaP,1, 2 suggesting that altered gene expression results in the elevated MIF protein detected in CaP. Recent studies have implicated tumor cell-derived MIF in the increased expression of macrophage-derived angiogenic activity.16 Those authors suggest that the interaction between tumor-infiltrating macrophages and tumor cells synergistically increases angiogenic potential. Together, these studies suggest that excessive MIF production may be associated with tumor progression and that, as such, increased MIF expression may serve as a biomarker for aggressive tumors.
To evaluate the utility of MIF as a prostate marker, we developed an enzyme-linked immunosorbent assay (ELISA) to determine MIF protein concentration in human serum samples.6 The current results reported here suggest that increased MIF concentration can be correlated with calculated free/total prostate specific antigen (f/t PSA) ratios. MIF often is elevated in serum from patients with a pathologic diagnosis of CaP. In addition, MIF concentrations are elevated significantly in serum from patients with tumors with a Gleason score ≥ 5 (analysis of variance [ANOVA]; P = 0.012). Relative quantitative reverse transcriptase-polymerase chain reaction (Q-RT-PCR) assays determined that MIF mRNA amounts within laser-capture microscopy (LCM)-dissected invasive CaP cells was 6.5 times greater than the amount detected in normal prostate epithelial cells from the same patient biopsy. The hypothesis to be tested is that increased MIF can serve as a biomarker for CaP.
MATERIALS AND METHODS
A prospective study was begun in November 2000 using serum samples obtained directly from the Bay Pines Veterans Administration Medical Center (VAMC) Radio Immune Assay (RIA) clinical laboratory. At the Bay Pines VAMC, all patients are screened routinely for serum PSA on an annual basis. New PSA assays were not conducted for this study. Rather, clinical PSA values obtained from the clinical laboratory of the Bay Pines VAMC by monoclonal RIA (Tandem-R; Hybritech, Inc. San Diego, CA) were used in this analysis. MIF concentrations were determined from serum aliquots of these clinical serum samples that were drawn for routine PSA screening. Five hundred nine serum samples that remained after routine clinical analysis for PSA were collected. Serum samples were assayed for MIF on the day they were obtained.
Approval for all facets of this study was obtained from the Bay Pines VAMC Institutional Review Board. These samples were obtained with waiver of informed consent under section 46.116(d) of the Department of Health and Human Services human subject regulations at 45 CFR 46 based on the following: 1) The research did not involve greater than minimal risk (samples are diagnostic specimens), 2) it was not practical to conduct this project without the waiver, 3) waiving informed consent would not adversely affect patients' rights and welfare (the diagnostic samples were obtained for reasons other than this research project), and 4) pertinent information would be provided to patients later if appropriate (once MIF validity as a prognostic marker is established, patients at risk would be notified of the research findings).
Serum MIF was quantified using a sandwich ELISA, as described previously.6, 17 Briefly, ELISA plates were prepared by coating wells with MIF monoclonal capture antibody (R&D Systems, Minneapolis, MN; monoclonal antibody 289) diluted to 2 μg/mL in sterile phosphate-buffered saline, pH 7.4. After an overnight incubation with capture antibody at 4 °C, plates were washed and then blocked (milk dilutent; catalog no. 500-82-01; Kirkegaard and Perry Laboratories, Gaithersburg, MD). A standard curve was generated using recombinant human MIF (no. 289-MF; R&D Systems) at concentrations from 2000 pg/mL to 31.25 pg/mL. For each patient sample, duplicate 100-μL serum samples were analyzed in three separate ELISA assays. MIF was detected using a biotinylated antihuman MIF antibody conjugate (1:250 dilution; catalog no. BAF289; R&D Systems), a strepavidin-alkaline phosphatase conjugate (1:1000 dilution; catalog no. 15-30-00; Kirkegaard and Perry Laboratories), and Blue Phos substrate (catalog no. 50-88-02; Kirkegaard and Perry Laboratories). Color development was allowed to proceed for 30 minutes, and an optical density (OD) of 630 nm was determined. Sigma Plot software (SPSS, Chicago, IL) was used for linear regression analysis. The equation from the line of best fit for the standards was used to determine serum MIF concentration. Undiluted serum samples with MIF concentrations > 1000 pg were reanalyzed by serial dilution until the OD 630 nm fell within the linear range of the standard curve.
Correlation between MIF, PSA, and TNM Classification
The computerized patient record system (CPRS) was used to collect clinical data, including corresponding serum PSA levels, f/t PSA ratios, TNM classification, and Gleason scores. Serum PSA values from the same clinical samples that were used to determine MIF were available for all patients in the study. The prostate pathology and Gleason score were used to group data to determine mean serum MIF concentrations in each category. We identified three pathologic groups: patients with benign prostatic hyperplasia (BPH), patients with CaP, and patients with high-grade prostate intraepithelial neoplasia (HGPIN). The CaP cohort (n = 61 patients) was stratified further according to documented Gleason score into three groups: patients with Gleason scores ≤ 4, patients with Gleason scores of 5 or 6, and patients with Gleason scores ≥ 7. Data from the CaP cohort were analyzed separately by stratifying according to the prostate tumor TNM classification18 as follows: T1c, clinically unapparent tumor not palpable or visible by imaging with the tumor identified by needle biopsy because of an elevated PSA level; T2a, tumor confined within the prostate with only one lobe involved; T2b, tumor confined within the prostate with both lobes involved; T3a, tumor extending through the prostatic capsule; T3b, tumor invading seminal vesicles; T4, tumor fixed or invading adjacent structures other than the seminal vesicles.
LCM-dissected glandular epithelial cells were obtained from the prostate tissue bank at the Histopathology Core at the H. Lee Moffitt Cancer Center. Tissue specimens in the bank had no identifying linkers and were from three different patients. The prostate tumor TNM classification for all three patients was pT3b. The Gleason score at the time of radical prostatectomy was documented as follows: Patient 1 had a Gleason score of 7 (4 + 3), Patient 2 had a Gleason score of 8 (4 + 4), and Patient 3 had a Gleason score of 6 (3 + 3). Dissected glandular epithelial cells (200 minimum) were isolated from a normal region, a cancerous region, and a cancerous region invasive to the margin within the same biopsy specimen. Staff pathologists at the Histopathology Core determined tissue pathology prior to LCM. Total RNA was isolated from these cells using the Paraffin Block RNA Isolation Kit (PBRNA; Ambion, Austin, TX; average yield, 500 ng total RNA). Q-RT-PCR analysis was performed using MIF specific primers and 18S rRNA primers, as described previously.5 Briefly, total RNA was reverse transcribed using Moloney murine leukemia virus reverse transcriptase at a concentration of 2 U/μg for 2 hours at 42 °C. PCR was carried out using ReadyMix Taq (Sigma, St. Louis, MO), MIF specific primers (upstream: 5′-CTCTCCGAGCTCACCCAGCAG-3′; downstream: 5′-CGCGTTCATGTCGTAATAGTT-3′) and a 2:8 competimer:primer 18S ratio (Alternate 18S Internal Standards; Ambion), as described previously.5 PCR products were separated using 2% agarose electrophoresis gels in 1 × 40 mM Tris-acetate plus 1 mM ethylenediamine tetraacetic acid and were stained with ethidium bromide. The area intensity of the 354-base pair (bp) 18S rRNA PCR product was used to standardize the area intensity of the MIF PCR product (254 bp). Data are expressed as the MIF expression ratio, which was defined as the MIF band area intensity divided by the 18S rRNA band area intensity ± standard deviation of three separate Q-RT-PCR reactions from three different patient biopsy samples.
Raw data were compiled and used to determine means, standard deviations, standard error of the means (SEMs), and ranges. Regression analyses, ANOVAs, and correlation coefficients were calculated using SigmaStat software (SPSS). Q-RT-PCR band intensity was quantified by direct area-intensity measurements on ethidium bromide-stained agarose gels using an ultraviolet transilluminator, a digital imaging system (AlphaImager 2200; AlphaInnotech, San Leandro, CA), and SigmaScan Pro image-analysis software (SPSS). Significance was determined at the α = 0.05 level, and all data are the mean ± SEM.
A total of 509 patient clinical serum samples drawn for PSA analysis were analyzed for MIF concentration. Fourteen samples were obtained from clinical serum samples for thyroid-stimulating hormone analysis and served as baseline controls. Prostate pathology was determined by a data base search. The CPRS search identified 61 patients with a histologically confirmed diagnosis of CaP with varying stages, Gleason scores, and tumor grades; 7 patients with pathologically confirmed HGPIN; and 84 patients with a histologic diagnosis of BPH. The data from the remaining 357 patients had no documented prostate pathology. The CaP cohort (n = 61 patients) was stratified further into three groups based on patients with Gleason scores ≤ 4 (n = 16 patients), Gleason scores of 5 or 6 (n = 24 patients), and Gleason scores ≥ 7 (n = 21 patients). Additional analyses were performed by stratifying this same CaP cohort (n = 61 patients) based on the prostate tumor TNM classification: T1c (n = 12 patients), T2a (n = 18 patients), T2b (n = 16 patients), T3a (n = 6 patients), T3b (n = 5 patients), and T4 (n = 4 patients).
Serum MIF Correlation with PSA
Preliminary regression analysis using data from all 509 patient samples determined a positive correlation between total PSA and serum MIF (correlation coefficient [r2] = 0.61; P < 0.001; n = 509 patients). This result indicates that PSA and MIF concentrations in serum tend to increase together. Values were clustered at PSA levels < 4 ng/mL and at MIF levels < 3 ng/mL. These studies established the utility of the serum MIF ELISA and suggest a correlation between serum MIF and PSA levels.
Correlation between MIF and f/t PSA in Serum
Calculated f/t PSA ratios are used to predict malignancy in patients with PSA levels in the 3.5–10 ng/mL ranges. Lower ratios usually are indicative of malignancy. Calculated f/t PSA ratios were available for 35 of 61 CaP patient serum samples. Best-fit polynomial regression analysis of MIF versus f/t PSA ratios were plotted to define a correlation coefficient of r2 = 0.88 (Fig. 1), indicating that, as f/t PSA ratios decrease in CaP patient sera, MIF levels increase.
Correlation between MIF and Prostate Pathology
Documented prostate pathology was used to group data into one of three identified groups: BPH (negative CaP prostate biopsy), CaP, and HGPIN. The mean serum MIF level was determined for each of these groups (Table 1). Clinical serum samples from patients that were used to analyze thyroid-stimulating hormone levels were used as controls (n = 14 samples). A pairwise Dunn multiple comparison test of a Kruskal–Wallis ANOVA on ranks identified significant differences in serum MIF levels between control and HGPIN samples and between control and CaP samples (P < 0.001). These data suggested a correlation between CaP and elevated serum MIF levels. A Spearman rank-order correlation was used to determine associations between PSA and MIF levels in patients with CaP. The variability in MIF concentration within this category failed to show a correlation unless the data were stratified according to PSA levels (Table 2). These data suggested that patients in the CaP category could be stratified into two groups (low PSA and high PSA). In general, patients with high PSA values (> 4 ng/mL) tend to have higher serum MIF concentration (r = 0.771; Spearman correlation coefficient), whereas patients with low PSA values (< 4 ng/mL) tend to have lower serum MIF concentration with a poorer correlation (r = −0.337; Spearman correlation coefficient). These data suggest that patients in the CaP cohort could be stratified into groups on the basis of both MIF concentration and PSA concentration. Four groups within the CaP cohort were defined using PSA > 4 ng/mL and MIF > 6 ng/mL as the high concentration parameters (Table 3). MIF > 6 ng/mL was defined as high serum MIF based on our data and that from two other laboratories that defined normal serum MIF as 4.0 ng/mL ± 2.3 ng/mL.17, 19 Student t tests defined a significant difference in the serum MIF mean values for the low PSA groups (Group 1 vs. Group 3) and the high PSA groups (Group 2 vs. Group 4) (P < 0.001), suggesting that four distinct groups based on the mean serum MIF and PSA values could be identified within the CaP cohort.
|Prostate pathology||Sample size (no.)||Serum MIF (mean ± SE)|
|Control||14||2.6 ± 1.34|
|BPH||84||2.7 ± 0.58|
|CaP||61||6.8 ± 0.87|
|HGPIN||7||10.1 ± 3.35|
|Total serum PSA (ng/mL)||Sample size||Correlation coefficient||P value|
|Group||Sample size||MIF (ng/mL)||Total PSA (ng/mL)||Description|
|1||29||2.4 ± 0.2||2.0 ± 0.2||Low PSA/low MIF|
|2||10||2.9 ± 0.3||8.3 ± 1.0||High PSA/low MIF|
|3||18||12.1 ± 1.7||0.5 ± 0.1||Low PSA/high MIF|
|4||4||10.2 ± 3.5||19.6 ± 9.4||High PSA/high MIF|
To determine whether serum MIF concentrations are correlated with prostate tumor TNM classification, we stratified the CaP cohort into groups based on prostate TNM classification and determined their mean serum MIF concentrations (ng/mL). In this analysis, the mean serum MIF concentrations showed no significant difference between the different prostate tumor classifications (ANOVA; P = 0.116) (Table 4). However, an ANOVA of the CaP cohort stratified according to Gleason score (≤ 4, 5–6, and > 7) determined a significant difference (P = 0.012) in mean serum MIF concentrations (ng/mL). A Dunn pairwise comparison determined a significant increase (P < 0.05) in serum MIF concentrations (ng/mL) in patients who had Gleason scores of 5–6 and ≥ 7 compared with patients who had Gleason scores ≤ 4. These data indicate that increased serum MIF concentrations are associated with higher Gleason grade tumors.
|Characteristic||Sample size (%)||Serum PSA (mean ng/mL)||Serum MIF (mean ng/mL)|
|T1c||12 (19.7)||5.13 ± 1.81||4.87 ± 1.18|
|T2a||18 (29.5)||2.01 ± 0.66||5.88 ± 0.66|
|T2b||16 (26.2)||3.27 ± 1.30||7.80 ± 0.67|
|T3a||6 (9.8)||13.73 ± 6.25||5.73 ± 1.20|
|T3b||5 (8.2)||0.83 ± 0.03||5.92 ± 0.65|
|T4||4 (6.6)||1.65 ± 1.02||13.57 ± 5.43|
|≤ 4||16 (26.2)||3.23 ± 2.11||2.13 ± 0.73|
|5 or 6||24 (39.3)||3.78 ± 1.20||6.44 ± 0.51|
|≥ 7||21 (34.4)||4.49 ± 1.59||8.67 ± 1.34|
MIF mRNA Expression in Invasive CaP Cells
Previous experiments have documented that in vitro cultured prostate cells express MIF mRNA and that MIF message amounts are significantly greater in CaP cell lines.5 However, increased MIF mRNA expression within CaP epithelial cells has not been documented in vivo. To associate altered serum MIF with MIF mRNA expression, LCM dissection of prostate epithelial cells was used to determine MIF mRNA amounts by Q-RT-PCR. Expression analysis of LCM-dissected cells excludes mRNA that may be contributed by stromal cells or immune cells. These studies looked at MIF expression in glandular epithelial cells only.
Figure 2 shows that MIF mRNA amounts, normalized to 18S rRNA, are greater in the cells that are invasive to the margin compared with normal and focal tumor cells from the same biopsy. Area-intensity measurements defined arbitrary MIF expression ratios (MIF area intensity/18S rRNA area intensity) for each of the prostate epithelial cell types. Arbitrary MIF mRNA expression ratios for each tissue category (normal, CaP, and invasive to margin) from three separate Q-RT-PCR analyses in three different patients (all with Stage pT3b tumors) were analyzed by using an ANOVA. The ANOVA defined significant differences (P < 0.001) between normal prostate epithelial cell MIF expression ratios (0.38 ± 0.06), focal CaP MIF expression ratios (1.88 ± 0.12), and invasive CaP expression ratios (2.48 ± 0.23). Tukey pairwise multiple comparisons determined that MIF expression ratios were significantly higher (P < 0.05) in CaP and invasive CaP prostate epithelial cells compared with normal epithelial cells from the same biopsy. Negative controls using RNA-free water instead of total RNA in the RT reaction mix did not show any amplified product, nor was a band visible after direct amplification of purified genomic DNA (data not shown). These results indicate that the increased amount of serum MIF detected in patients with CaP may be due in part to increased amounts of MIF mRNA produced by CaP epithelial cells. These results document elevated in vivo MIF mRNA expression in CaP.
The utility of MIF as a prognostic marker is suggested by the finding that the mean serum MIF concentration often is greater in patients with PSA levels < 2 ng/mL and a positive diagnosis of CaP (Table 3). In addition, patients who have tumors with a higher Gleason score (≥ 5) exhibit significantly increased serum MIF concentrations (Table 4). Many different tumor cells secrete MIF, and MIF has been associated with angiogenesis.2, 5, 10, 16 These findings suggest that this cytokine promotes tumor survival by inducing an angiogenic response, but MIF by itself is not directly angiogenic. Recently, White and coworkers demonstrated that MIF produced by nonsmall cell lung carcinoma cells induced angiogenic cytokine expression by macrophages.16 These recent data, along with the results of our previous studies, suggest a hypothesis whereby increased MIF secretion by tumor cells aids in tumor promotion, survival, and metastasis by inducing the release of angiogenic factors by surrounding nontumor cells.
The exact mechanism by which tumor cells up-regulate MIF expression is unknown. We demonstrated that, within microdissected CaP cells, mRNA expression was highest in cells invasive to the prostate margin compared with histologically normal epithelial cells (Fig. 2). Based on our previous MIF mRNA stability study results utilizing in vitro cultured DU-145 CaP epithelial cells,5 this may be due in part to increased message stability in tumor cells. This study also demonstrated that MIF secreted by in vitro cultured CaP cells is more stable than that secreted by normal prostate epithelial cells.5 Increased serum MIF concentrations were observed in patients with CaP, irrespective of treatment modality, indicating that continued MIF secretion by the CaP epithelial cell may not be regulated hormonally. This finding is supported by a study correlating MIF immunostaining intensity with Gleason grade and combined endocrine treatment.20 Combined with our previous findings of enhanced MIF stability and secretion by CaP cells,5 our current results and those of other groups7, 20 suggest that elevated serum MIF detected in patients with CaP is derived from CaP cells.
CaP cells that over-express the MIF gene product exhibit an increased potential to induce angiogenesis, thereby exhibiting increased survival and potential to metastasize. Therefore, determining either the patient serum MIF concentration or the amount of MIF expressed in a prostate biopsy specimen may help predict aggressive disease and may prove useful as a new prognostic marker. Should this hypothesis prove correct, monitoring patients with CaP for serum MIF concentration on a routine basis may be used for stratifying patients according to their risk for metastasis. New prognostic test development for CaP is vitally important, because, although much is understood about the relation between PSA concentration and growth in the early stages of CaP, as the disease disseminates, the relation between tumor burden and PSA concentration becomes less well defined.
In summary, the combined assessment of serum total PSA and MIF concentrations in patients with CaP has identified four patient groups. These data indicate that patients with CaP who have low serum total PSA levels (< 4 ng/mL) and high serum MIF levels (> 6 ng/mL) potentially have a higher risk of recurrence compared with patients who have low serum total PSA levels and low serum MIF levels (< 6 ng/mL). This preliminary study demonstrates that MIF is a biomarker for CaP and deserves further evaluation in a large population to determine whether it can aid in the identification and follow-up of patients with recurrent disease.
The authors thank William Webster, Pharm.D., for his critical review of the article.
- 14An essential role for macrophage migration inhibitory factor (MIF) in angiogenesis and the growth of a murine lymphoma. Mol Med. 1999; 5: 161–191., , , , , .
- 17Quantitation of macrophage migration inhibitory factor (MIF) using the one-step sandwich enzyme immunosorbent assay: elevated serum MIF concentrations in patients with autoimmune diseases and identification of MIF in erythrocytes. Int J Mol Med. 2000; 5: 397–403., , , et al.
- 18American Joint Committee on Cancer. Prostate. AJCC cancer staging manual. Philadelphia: Lippincott-Raven, 1997: 219–222.