| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Original Studies |
The Intramural Research Program, National Institute on Aging, National Institutes of Health (S.M.H., E.J.M.), Baltimore, Maryland 21224; and Departments of Medicine (M.R.B.) and Urology (P.K.L., H.B.C.), The Johns Hopkins University School of Medicine, Baltimore, Maryland 21287
Address all correspondence and requests for reprints to: S. Mitchell Harman, M.D., Ph.D., Kronos Research Foundation, 4455 East Camelback Road, Suite B135, Phoenix, Arizona 85018. E-mail: harman{at}thekronosgroup.com
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The importance of the IGF system in prostate growth is underscored by the detection of every element of this system, including IGF type 1 receptors (10, 11, 12), IGF-I and IGF-II messenger ribonucleic acids (mRNAs) (10, 12, 13, 14, 15, 16, 17), and IGF-I (18) and IGF-II (14, 19) protein, as well as IGFBP2- through 6 (10, 15) in normal, hyperplastic, and/or neoplastic prostate cells and tissues. The activity of this system in the prostate is suggested by the findings that both IGF-I and IGF-II stimulate prostate cell growth (20), blocking the action of IGF inhibits prostate cell proliferation (18, 21), IGFBP-3 blunts the proliferative effect of IGF-I or IGF-II on prostate cells (20), and, finally, prostate-specific antigen (PSA) functions as an IGFBP protease (22), which reverses the inhibitory effect of IGFBP-3 on IGF-I action in vitro (20).
Given the above, it is reasonable to ask whether circulating IGFs may play a role in the induction or promotion of prostate cancer and/or hyperplasia. Based on a study of blood samples drawn from men who later developed prostate cancer and age-matched men who did not, Chan et al. (23) stated that the association between circulating IGF-I level and risk of prostate cancer was stronger than that of any previously reported risk factor. However, various prior investigations have found serum PSA to be powerfully predictive of prostate cancer (24, 25, 26, 27, 28, 29, 30). Although Chan et al. showed IGF-I to remain a significant predictor of prostate cancer in men with low (<4.0 ng/mL) and high (>4.0 ng/mL) PSA levels, they did not directly compare the predictive power of plasma PSA with that of IGF-I. Nonetheless, their report has already been cited in a subsequent publication as demonstrating, "... that plasma IGF-I concentration may be a better predictor of prostate cancer than PSA" (31). Another question is whether the reported association of high IGF-I levels and prostate cancer could represent ascertainment bias. If high IGF-I were to be associated with greater prostate size (but not carcinogenesis), this condition could more frequently bring men with high IGF-I levels to the attention of urologists, resulting in a higher rate of diagnosis of prostate cancer.
In the present study, using stored sera from men followed in the Baltimore Longitudinal Study on Aging (BLSA), we investigated whether the circulating IGF-I level is an independent predictor of prostate cancer and compared its predictive value with those of IGF-II, IGFBP-3, and PSA. We also examined the associations between IGF-I, IGF-II, IGFBP-3, and PSA and prostate size.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
All subjects were male participants in the BLSA, a largely middle-class, 87% Caucasian population, whose characteristics have been described previously (32). The BLSA, an open registration study of the physiology of aging, has for more than 40 yr accumulated data on men studied at approximately 2.5-yr intervals at visits to the NIA’s Gerontology Research Center in Baltimore, MD. The BLSA investigative protocol is approved by the combined institutional review boards of the Johns Hopkins Bayview Medical Center and the Gerontology Research Center. All subjects studied signed institutional review board-approved informed consent documents. Each participant receives an extensive interim medical and psychological examination at each visit, and serum and plasma samples are banked for future investigation. Since 1992, prostate evaluations have included PSA measurements.
We identified 120 BLSA participants diagnosed with prostate cancer between 1960 and 1997. Diagnosis was histologically confirmed in 96 cases and by clinical history and chart review in the remaining 24. Banked sera from a visit (hereafter designated the index visit) nearest to 10 yr before diagnosis (range, 3.1–16.2; mean ± SD, 9.2 ± 2.4; median, 10.0 yr) were available for 85 of these men (cases). We attempted to match 2 controls to each case, but suitable sera were available from only 146 of the 170 potential controls initially identified, so that there was only a single control available for 24 of the cases. Control subjects were BLSA participants with an index visit at the same age (±2 yr) as the corresponding case and a similar length of follow-up free of the diagnosis of prostate cancer (range, 3.7–17.1; mean ± SD, 9.3 ± 2.1 yr; median, 10.0 yr). The mean difference between cases and their individual age-matched controls for length of follow up was 0.10 ± 2.7 (±SD) yr with a maximum difference of +7.4 yr (i.e. 7.4 yr more follow-up for the control than the case). To strengthen the analyses of prostate size, we added an additional cohort of 85 BLSA men (normal subjects) without prostate cancer (follow-up range, 3.1–16.0; mean, 9.9; median, 10.0 yr) in whom prostate volume by magnetic resonance imaging (MRI) was known, but who were not age matched to cases.
Samples
Serum samples were stored at -80 C. Because sample storage time varied greatly (range, 3.4–34; mean, 12.8 yr), we measured serum sodium by routine clinical laboratory methodology (sodium electrode) to detect evaporation or dilution of samples. Serum sodium was elevated (>150 mmol/L) in 22 and low (<130 mmol/L) in 19 of the samples examined, nearly evenly distributed among the 3 study groups. The high sodium values were distributed across the range of visit dates from 1975 to 1994 more or less in proportion to the number of samples obtained in each 5-yr period. In contrast, the low sodium samples tended to be concentrated between 1985 and 1990, with 45% of them in a single year (1988). To avoid confounding effects of sample artifacts, all samples with low or elevated sodium concentrations, as defined above, were excluded from further analyses, reducing the numbers of sera finally available for analysis to 72 cases, 127 controls, and 76 normal subjects. Of the 72 cases, 15 were in stage T1 (5 a, 3 b, and 7 c), 11 were in stage T2 (1 unmodified, 5 a, and 5 b), 3 were in stage T3, 7 had advanced disease/metastasis, and in 36 staging was unknown.
Assays
Samples were thawed, aliquoted, refrozen, and sent to Endocrine Sciences, Inc. (Calabasas Hills, CA), to be assayed for IGF-I, IGF-II, and IGFBP-3. IGF-I was measured using an IGF-II-blocked RIA in which, in the presence of an antibody highly specific for IGF-I, excess IGF-II was added to eliminate any influence of residual IGFBPs present after acid-ethanol extraction (33). The minimum detectable dose (MDD) for this assay averaged 10 µg/L, with intraassay coefficients of variance (CV = SD/mean x 100) of 20%, 5.6%, and 4.6% at concentrations of 26, 330, and 603 µg/L, respectively, and interassay CVs of 9.0%, 9.9%, and 10% at concentrations of 65, 118, and 403 µg/L. IGF-II was assayed using an IGF-I blocked RIA after acid-ethanol extraction with an antibody highly specific for IGF-II, with excess IGF-I added to eliminate the influence of residual IGFBPs (34). The MDD for this assay was 20 µg/L, with intraassay CVs of 9.6%, 7.1%, and 4.9% at concentrations of 91, 380, and 613 µg/L, and interassay CVs of 30%, 9.0%, and 5.1% at concentrations of 82, 384, and 568 µg/L. IGFBP-3 was measured by direct RIA without extraction (35), with a MDD of 0.3 mg/L and intraassay CVs of 13% and 5.1% at concentrations of 1.0 and 2.7 mg/L and interassay CVs of 17% and 5.5% at concentrations of 0.8 and 2.9 mg/L. PSA was determined in the clinical research laboratory of the Johns Hopkins Hospital by monoclonal RIA (Tandem-R, Hybritech, Inc., San Diego, CA). New PSA assays were not conducted for the purpose of this study. Rather, PSA values previously determined from serum aliquots drawn at the same visit were employed (201 subjects). Where no simultaneous PSA value existed, a PSA value was interpolated from multiple values measured at visits before and/or after the index visit, or, if only a single PSA measurement was available, this value was assigned, provided it was obtained fewer than 5 yr before or after the index visit (67 subjects). However, for 7 subjects, no PSA determination was made within 5 yr of the index visit. Thus, we were able to assign PSA values to 268 of the 275 men studied.
Prostate volume measurements
Prostate volume was determined by MRI using a dedicated phased array pelvic coil (1.5 T). Measurements were made in centimeters in three planes (axial, sagittal, and coronal), and prostate volume was calculated in cubic centimeters according to the formula; volume = 4/3 [(axial value/2)(coronal value/2)(sagittal value/2)]. No MRI measurements were available for cases. On the average, MRI measurements were obtained 8.6 ± 2.1 (mean ± SD) yr (range, 2.3–15.1) after the index visit in controls and 8.8 ± 2.3 yr (range, 1.9–13.9) after the index visit in normal subjects. Therefore, like prostate cancer, prostate volume was a prospective variable, with levels of IGF-I, IGF-II, IGFBP-3, and PSA examined with regard to how they related to prostate volumes measured 8–9 yr later.
Statistical methods
Raw data were compiled in Excel (Microsoft Corp.) spreadsheets. Means, SDs, and ranges were obtained, and regressions and analyses of covariance were performed using StatView 5.0 for Macintosh (SAS Institute, Inc., Cary, NC). Each of the variables measured was examined for the effects of age and length of sample storage (i.e. date) by linear regression against subject age at and date of the index visit. For the purpose of these calculations, we used a simple computer algorithm to express date as a decimal number to the nearest tenth of a year (e.g. 7/1/93 = 1993.5). Correlation coefficients (Pearson’s r) and P values for significance of slope were examined. Values for subject characteristics were compared between groups (cases and controls) by ANOVA.
We ranked subjects in pooled case and control groups into low, mid, and high tertiles for each of the continuous variables measured and performed conditional logistic regression analysis (36) using SPSS, version 8 (SPSS, Inc., Chicago, IL), to estimate risk as the odds ratios (OR) for the development of prostate cancer. Independent variables were examined as tertiles to give levels of exposure. In these analyses, the mid and high tertiles were compared with the low tertile. Because OR estimates are skewed by the overrepresentation of cases in case-control studies compared with randomly selected populations, they do not correspond to the true population risk ratios (RR). Therefore, for each significant adjusted OR, we estimated the corresponding RR using the method proposed by Zhang et al. (37).
To assess the potential diagnostic power of variables found to be significant predictors of prostate cancer (PSA, IGF-I, and IGF-II), we calculated the operating characteristics, specificity, and sensitivity (38) for each variable and plotted receiver operator curves. To examine pairs of variables, we plotted a series of receiver operator curves for PSA with IGF-I or IGF-II at each of a range of levels of PSA. We constructed these sets of curves using each of two different algorithms: the first in which subjects were identified as positive if they met criteria for either variable ("or" algorithm) and the second in which they were required to meet criteria for both variables ("and" algorithm). Finally, to examine combinations of all three variables, we created 130 receiver operator curves (6 levels of IGF-I with each of 5 levels of IGF-II at each of 5 levels of PSA) using the "or" algorithm and then a set of similar curves using the "and" algorithm.
Adjustment for potential confounding factors
There was a pronounced effect of date of sample on IGF-II
(r = 0.337; P < 0.0001) with lower assayable
IGF-II levels in older samples, suggesting a systematic, gradual
deterioration of the IGF-II molecule or the epitope thereof measured by
the antibody employed in our RIA, with an average loss of 10.7 µg/L
(
2.4%)·yr over the 31-yr (1964–1995) period of sampling.
Although only a minority of the 275 samples assayed were more than 10
yr old (5.8% collected from 1963–1974 and 24.4% from 1975–1984), we
nonetheless adjusted IGF-II levels for sample date, as described below,
using a second order polynomial equation (date and
date2) which significantly improved the
relationship. No significant date effect was seen for IGF-I or IGFBP-3
assays. There were decreases with subject age in serum IGF-I (r =
-0.206; P < 0.005) and IGFBP-3 (r = -0.250;
P < 0.0005), but no significant effect of age on
IGF-II. PSA increased significantly with subject age (r = 0.299;
P < 0.0001). Adding an age2 term
did not improve the age relationships. To correctly assign tertiles, we
first used the regression equation for each variable’s relationships
to age and date to calculate, for each subject’s value of that
variable, a factor to adjust it to a standard age (60.8 yr) and date
(1994.6). We also employed date-adjusted values in the estimation of
operating characteristics. To adjust logistic regressions, we used
uncorrected values and included age, visit date, and
date2 as covariates in the analyses. To adjust
for covariation of measured variables, we repeated the logistic
regressions with other relevant measures included as independent
variables. To explore the relationship of prostate size to the serum
factors measured, we performed simple and multiple linear regressions
of age- and date-adjusted variables with correlation coefficients.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
As noted in Table 1
, cases and
controls were matched for age to within less than a year. However, mean
index visit dates differed between controls and cases by about 6 yr
(P < 0.0001), whereas there was no difference between
groups in mean duration of follow-up. Cases and controls did not differ
significantly with regard to weight or height, but cases had slightly,
but not significantly, lower body mass indices (calculated as
weight/height2), a measure of obesity.
Normal subjects were significantly younger than both cases and controls
(P < 0.0001) and were sampled more recently than the
cases (P < 0.0001). In addition, normal subjects were
heavier (P < 0 0.005) and taller (P <
0.05) than cases and taller (P < 0.001) than
controls.
|
Table 2 shows correlations of the
variables with one another, adjusted for effects of age, sample date,
date2, and diagnostic group (which had a
significant additional effect on these relationships). Even after
adjustment, concentrations of IGF-I, IGF-II, and IGFBP-3 covaried
closely with one another (P < 0.0001), but not with
PSA concentration.
|
Figure 1 shows the percentage of
cases and controls in each tertile. For IGF-I, a reciprocal
distribution of controls and cases with, respectively, progressively
diminishing and increasing fractions in the low, mid, and high tertiles
was apparent. An opposite trend of greater magnitude existed for
IGF-II. No progression of the IGFBP-3 distribution was apparent.
Finally, there were far greater proportions of controls in the low, and
of cases in the high, PSA tertiles, and there were similar percentages
of cases and controls in the mid tertile.
|
(upper panel)
demonstrates the results of single logistic regression of four
variables (IGF-I, IGF-II, IGFBP-3, and PSA) by tertile to give ORs and
95% confidence intervals (CIs) associated with each tertile. The high
tertile for IGF-I carried a positive OR of 1.65, but a CI that included
zero (i.e. nonsignificant). In contrast, IGF-II appeared
significantly protective with an OR for the high tertile of 0.24 and a
progressive protective trend from the low to the high tertile
(P < 0.01). High tertile IGFBP-3 also appeared
protective (OR = 0.71), but was not significantly so. Serum PSA
was the best predictor of prostate cancer, with significant positive
risk associated with membership in both mid (OR = 5.12) and high
(OR = 12.5) tertiles and a highly significant trend
(P < 0.0001).
|
, lower
panel), the OR associated with the high tertile for IGF-I was
amplified to 3.11 (CI, 1.11–8.74), which was significant. IGF-II
showed about the same magnitude of protective effect as previously
(OR = 0.20) and remained significant (CI, 0.07–0.59). The OR
associated with IGFBP-3 remained nonsignificant. The ORs for PSA were
essentially unchanged by adjustment for IGF-I, IGF-II, and IGFBP-3, at
12.5 for the high tertile and 4.95 for the mid tertile, both highly
significant. Estimates of the RRs and their corresponding CIs adjusted
for the other variables were: for high tertile IGF-I, 2.03 (CI,
1.08–2.95); for high tertile IGF-II, 0.31 (CI, 0.12–0.73); for high
tertile PSA, 5.54 (CI, 2.92–7.62); and for mid tertile PSA, 3.46 (CI,
1.46–4.73). Predictive value of measured variables
Figure 2A demonstrates receiver
operator curves for IGF-I, IGF-II, and PSA considered separately. The
curves shown were constructed by calculating the specificity and
sensitivity for prostate cancer in the study population at levels of
PSA varying in increments of 0.4 from 0.2–6.2 ng/mL, levels of IGF-I
in increments of 10 from 100–250 µg/L, and levels of
IGF-II in increments of 20 from 200–500 µg/L. For PSA and IGF-I,
the higher the serum level, the higher the sensitivity. For IGF-II the
opposite is true, with high sensitivity at low serum concentrations. At
very low sensitivities (<0.3; i.e. <30% of cases
classified as positive), the use of IGF-II alone provides a specificity
as good or better than that of PSA. The use of PSA provides better
specificity than IGF-II at sensitivity levels greater than 0.3.
Moreover, at every point PSA gives better specificity for a given
sensitivity (and vice versa) than does IGF-I. Figure 2B
shows a subset of the operating curves obtained for PSA and IGF-II
using the "or" algorithm. Sensitivities and specificities at levels
of IGF-II varying from 200–500 µg/L in increments of 20 µg/dL are
shown for PSA levels of 1.0, 1.4, 2.6, 3.0, 3.8, and 4.2 ng/dL. We
found that classifying subjects with PSA levels greater than 3.8 ng/dL
or an IGF-II less than 300 µg/L as positive resulted in a specificity
of 0.93 with a sensitivity of 0.53 (Fig. 2B, arrow). No
other algorithm or pair of variables did as well (data not shown).
Moreover, no combination of all three variables, PSA, IGF-II, and
IGF-I, gave results better than those for PSA and IGF-II (data not
shown).
|
As shown in Fig. 3, in the 147
control and normal subjects who had MRI measurements, IGF-I was not
significantly correlated with prostate size. In contrast, IGF-II was
significantly inversely related (r = -0.219; P <
0.01), and PSA was significantly directly related to prostate volume
(r = 0.461; P < 0.0001). IGFBP-3 was not
significantly related to prostate size. Not shown in the figure,
prostate size also increased with subject age (r = 0.223;
P < 0.01). Because IGF-I, IGF-II, and IGFBP-3 covary,
and because IGF-I, IGFBP-3, and, to a lesser extent, IGF-II are
affected by age, we performed a multiple regression to examine the
separate effects of age (r = 0.133; P > 0.1),
IGF-I (r = 0.024; P = NS), IGF-II (r = 0.245;
P < 0.003), IGFBP-3 (r = 0.156; P
> 0.1), and PSA (r = 0.377; P < 0.0001) on
prostate volume. The partial correlation coefficients confirm the
independent inverse relationship of IGF-II and the direct relationship
of PSA with prostate volume, whereas IGFBP-3, IGF-I, and age drop out
of the equation.
|
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The results of 6 of the 7 recent reports relating circulating IGF-I and related factors to human prostate cancer have recently been reviewed (39). Five of these studies (40, 41, 42, 43, 44) employed retrospective case-control designs, examining blood samples from patients already diagnosed with prostate cancer, some of whom had metastatic disease. Only 2 were, like ours, prospective in design, of which 1 with 269 cases (23) showed significant excess risks (OR = 4.3; P < 0.01) of prostate cancer with high vs. low quartile IGF-I, whereas the other, with 45 cases (45), did not (OR = 0.81; P > 0.7). In 1 case-control study with 158 cases (44), IGF-I levels were not predictive of prostate cancer (OR = 1.43; P = NS), whereas in a similar study of 160 cases (43), a marginally significant linear increase in risk of 1.91-fold/60 µg/L IGF-I (P = 0.05) was observed. In the 3 studies in which OR was not estimated, mean IGF-I was reported to be significantly increased in the cases in 1 (41), but not the other 2 (40, 42). Thus, our finding of a marginally significantly increased risk of prostate cancer associated with higher serum IGF-I levels parallels the results of prior studies.
As noted by Cohen (39), the standard acid-ethanol extraction technique used to prevent interference of IGFBP-3 fails to eliminate IGFBP-2 and -4. Therefore, IGF-I levels in the cases in the study with the greatest excess IGF-I-related risk (43) could have been artifactually elevated due to the known increases in IGFBP-2 and IGFBP-4 in patients with prostate cancer (40, 41, 42). However, this confound would be unlikely to affect the assays employed in the other two studies in which IGF-I was higher in cases (23, 41) or in the IGF-blocked assays in our study. Thus, assay artifact is probably not responsible for most reported associations of prostate cancer and serum IGF-I.
Our study also suggests that ascertainment bias was not responsible for the association of IGF-I and prostate cancer. First, we did not confirm the proposed relationship between IGF-I and prostate size, making the hypothesis that larger prostates in high IGF-I individuals might result in an increased rate of detection of prostate cancer unlikely. Second, the BLSA population undergoes prostate examinations and PSA measurements at regular intervals. Therefore, ascertainment bias would be far less likely in this population, in whom, nonetheless, we observed a significant relationship of serum IGF-I to risk of prostate cancer.
With regard to IGFBP-3, Chan et al. (23) reported a significant reduction in cancer risk (OR = 0.33) only after adjustment for effects of IGF-I. This protective effect was found in the third, but not the fourth (OR = 0.41), quartile for IGFBP-3 levels, which gave a nonsignificant progression by trend analysis (P = 0.09). The negative risk with high IGFBP-3 in the present study was of lesser magnitude and nonsignificant, as previously reported by others (44). Thus, the role, if any, of IGFBP-3 in modulating the effects of IGF-I on prostate cancer risk remains unclear.
A unique finding in the present study was the significant reduction in
risk of prostate cancer associated with high tertile IGF-II. Consistent
with the possibility of a protective effect of IGF-II, is our
additional observation that serum IGF-II level was inversely associated
with prostate volume by MRI. Prior investigators have reported
significantly lower IGF-II levels (38%; P < 0.03) in
prostate cancer cases vs. controls in one study
(42) and nonsignificant decreases of 7% (40)
and 8% (41) in two others. Only Chan et al.
(23) have previously estimated relative risks of prostate
cancer associated with IGF-II levels, reporting no apparent IGF-II
effect (OR = 0.97). This disparity between their study and ours
may reflect methodological differences in the IGF-II assays, variation
in the time at which subjects were studied relative to the development
of cancer, or other unknown factors. In a recent study of a prostate
cancer model, the TRAMP transgenic mouse, IGF-I mRNA in the prostate as
well as serum IGF-I increased as the tumor progressed, whereas IGF-II
expression was 80% reduced in TRAMP compared with control mice
(31), suggesting a reciprocal relationship between IGF-I
and IGF-II actions in prostate cancer. The mannose-6-phosphate IGF-II
receptor is a potent activator of transforming growth factor-
, which
is a growth inhibitor for many cell types. Recent findings in breast
cancer cells of loss of heterozygosity and mutations at the
mannose-6-phosphate IGF-II receptor gene locus suggest that the
normal receptor functions as a tumor suppressor (46, 47).
IGF-II receptor mRNA has been detected in cultured prostate cancer
cells in one study (16), but not another
(10). The above findings taken together with the present
data suggest that IGF-II could reduce prostate growth and oppose
prostate cancer development via actions at the IGF-II receptor.
PSA is a protease in the kallikrein family (48), secreted by normal and malignant prostate epithelial cells (40, 49, 50). PSA is produced in smaller quantities in a wide variety of tissues other than prostate (30). The serum PSA concentration is a strong predictor of prostate cancer (24, 25, 26, 27, 28, 29, 30). In the present study, distributions of PSA were skewed toward higher values in the cases, and both high and mid PSA tertiles were associated with greater relative risks of cancer than any other tested variable. Although PSA levels of 4 ng/mL or greater are generally accepted as suspicious, Gann et al. (27) reported that men with PSA levels from 2–3 ng/mL have a 4-fold increased risk of prostate cancer developing in the subsequent 10 yr, and that men with values from 1–1.5 ng/mL have double the risk, compared with those having PSA values less than 1 ng/mL. These estimates are very similar to those in the current report.
Of the 72 cases we studied, 14 (20%) had prostate cancer diagnosed
less than 8 yr (range, 3.1–7.9 yr) after the index visit, because the
index visit was their first BLSA visit (i.e. no samples more
remote from the time of diagnosis were available). It may be argued
that these men had high PSA levels because they already had prostate
cancer at the time of sampling (although it was not detected on
clinical examination). In fact, the mean PSA level in the case group
decreases from 3.3 to 2.6 ng/mL when these 14 men are excluded, a value
still highly significantly different from that for controls
(P < 0.0001). Given the finding of a high incidence of
microfoci of cancer in prostate glands at autopsy of young and
middle-aged men dying of unrelated causes (51), it is
likely that even subjects diagnosed 10 yr or more after the index visit
already had subclinical cancer at the time of sampling. Thus, the 14
men in whom the diagnosis of cancer occurred relatively soon after
sampling differed only quantitatively, rather than qualitatively, from
the remainder of the cases.
The relationships reported between circulating IGF-I, IGF-II, and possibly IGFBP-3 and prostate cancer carry both diagnostic and therapeutic implications. First, with regard to early diagnosis, it is important to note that ORs, per se, are not very indicative of the utility of a biomarker (52), a point illustrated by our finding that IGF-I, despite having a significant OR for prostate cancer, provides relatively poor specificity and sensitivity, as estimated from receiver operator curves. The level of specificity that we observed for PSA is comparable with that reported previously in more than 22,000 men followed prospectively for up to 10 yr (27) and confirms that even a single PSA measurement is a powerful tool for estimating the risk of prostate cancer. In a study of 665 men, 179 of whom had prostate cancer (53), published after submission of the current report, the investigators concluded that in men with elevated serum PSA, serum IGF-I is not a useful diagnostic test for prostate cancer, a conclusion entirely congruent with our own. Our data suggest that the assessment of IGF-II (but not IGF-I) in addition to PSA can improve sensitivity and/or specificity. Confirmation of this finding would require further research on larger populations in which the distribution ratio of cases to controls mirrors that in the population at large.
With regard to therapy, although observational studies such as the current report and that by Chan et al. (23) cannot establish whether IGF-I plays an etiological role in prostate cancer, if IGFs truly accelerate prostate cancer growth, then treatments that may elevate intraprostatic IGF-I, such as GH replacement, should probably be avoided in men at increased risk for or with known or suspect prostate cancer. In addition, strategies to reduce the levels or block the effects of IGFs in the prostate may have value. Measurement of longitudinal changes in growth factors over time in cases and controls might shed further light on this issue.
In summary, the current study is in agreement with the conclusion of Chan et al. (23) that circulating IGF-I is a statistical risk factor for clinical prostate carcinoma in the subsequent 5–10 yr. Whether plasma IGF-I is an etiological factor or is, like PSA, a marker for cancer cannot be deduced from the current data. However, the association of IGF-I and prostate cancer is probably not the result of ascertainment bias. Our finding that higher IGF-II levels are associated with a reduced likelihood of prostate cancer and smaller prostate size requires further investigation. Although PSA is a far more powerful predictor of prostate cancer than either IGF-I or IGF-II, IGF-II measurement could add significantly to the diagnostic accuracy of PSA. We conclude that the role of IGFs in both etiology and diagnosis of prostate cancer should be further characterized prospectively in larger populations, in which the influences of age, ethnicity, rates of change, and other factors can be fully evaluated.
|
Acknowledgments |
|---|
Received November 19, 1999.
Revised February 21, 2000.
Accepted August 21, 2000.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
-reductase and steroid hormone receptor messenger
ribonucleic acid (mRNA) levels, but not in insulin-like growth factor
mRNA levels. J Clin Endocrinol Metab. 77:1203–1208.[Abstract]
This article has been cited by other articles:
 |
|
 |
|
  A. W. Roddam, N. E. Allen, P. Appleby, T. J. Key, L. Ferrucci, H. B. Carter, E. J. Metter, C. Chen, N. S. Weiss, A. Fitzpatrick, et al. Insulin-like Growth Factors, Their Binding Proteins, and Prostate Cancer Risk: Analysis of Individual Patient Data from 12 Prospective Studies Ann Intern Med, October 7, 2008; 149(7): 461 - 471. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Anzo, L. J. Cobb, D. L. Hwang, H. Mehta, J. W. Said, S. Yakar, D. LeRoith, and P. Cohen Targeted Deletion of Hepatic Igf1 in TRAMP Mice Leads to Dramatic Alterations in the Circulating Insulin-Like Growth Factor Axis but Does Not Reduce Tumor Progression Cancer Res., May 1, 2008; 68(9): 3342 - 3349. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. A. Darbinian, A. M. Ferrara, S. K. Van Den Eeden, C. P. Quesenberry Jr., B. Fireman, and L. A. Habel Glycemic Status and Risk of Prostate Cancer Cancer Epidemiol. Biomarkers Prev., March 1, 2008; 17(3): 628 - 635. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Z. Wang, R. M. Luque, R. D. Kineman, V. H. Ray, K. T. Christov, D. D. Lantvit, T. Shirai, S. Hedayat, T. G. Unterman, M. C. Bosland, et al. Disruption of Growth Hormone Signaling Retards Prostate Carcinogenesis in the Probasin/TAg Rat Endocrinology, March 1, 2008; 149(3): 1366 - 1376. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. H. Siahpush, T. L. Vaughan, J. N. Lampe, R. Freeman, S. Lewis, R. D. Odze, P. L. Blount, K. Ayub, P. S. Rabinovitch, B. J. Reid, et al. Longitudinal Study of Insulin-like Growth Factor, Insulin-like Growth Factor Binding Protein-3, and their Polymorphisms: Risk of Neoplastic Progression in Barrett's Esophagus Cancer Epidemiol. Biomarkers Prev., November 1, 2007; 16(11): 2387 - 2395. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J.-R. Zhou, L. Li, and W. Pan Dietary soy and tea combinations for prevention of breast and prostate cancers by targeting metabolic syndrome elements in mice Am. J. Clinical Nutrition, September 1, 2007; 86(3): 882S - 888S. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. E. Allen, T. J. Key, P. N. Appleby, R. C. Travis, A. W. Roddam, S. Rinaldi, L. Egevad, S. Rohrmann, J. Linseisen, T. Pischon, et al. Serum Insulin-like Growth Factor (IGF)-I and IGF-Binding Protein-3 Concentrations and Prostate Cancer Risk: Results from the European Prospective Investigation into Cancer and Nutrition Cancer Epidemiol. Biomarkers Prev., June 1, 2007; 16(6): 1121 - 1127. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. U. Garcia, K. Urbanska, L. Koltowski, K. Reiss, and C. Sell Insulin-Like Growth Factor-I Produced by Seminal Vesicles: Relationship to Intraepithelial Basal Cell Hyperplasia in the Prostate Clin. Cancer Res., June 1, 2007; 13(11): 3140 - 3146. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Rodriguez, S. J. Freedland, A. Deka, E. J. Jacobs, M. L. McCullough, A. V. Patel, M. J. Thun, and E. E. Calle Body Mass Index, Weight Change, and Risk of Prostate Cancer in the Cancer Prevention Study II Nutrition Cohort Cancer Epidemiol. Biomarkers Prev., January 1, 2007; 16(1): 63 - 69. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. D. Tran, C. Diorio, S. Berube, M. Pollak, and J. Brisson Relation of insulin-like growth factor (IGF) I and IGF-binding protein 3 concentrations with intakes of fruit, vegetables, and antioxidants Am. J. Clinical Nutrition, December 1, 2006; 84(6): 1518 - 1526. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Chen, R. Freeman, L. F. Voigt, A. Fitzpatrick, S. R. Plymate, and N. S. Weiss Prostate Cancer Risk in Relation to Selected Genetic Polymorphisms in Insulin-like Growth Factor-I, Insulin-like Growth Factor Binding Protein-3, and Insulin-like Growth Factor-I Receptor Cancer Epidemiol. Biomarkers Prev., December 1, 2006; 15(12): 2461 - 2466. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. P Singh and R. Agarwal Mechanisms of action of novel agents for prostate cancer chemoprevention. Endocr. Relat. Cancer, September 1, 2006; 13(3): 751 - 778. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Severi, H. A. Morris, R. J. MacInnis, D. R. English, W. D. Tilley, J. L. Hopper, P. Boyle, and G. G. Giles Circulating insulin-like growth factor-I and binding protein-3 and risk of prostate cancer. Cancer Epidemiol. Biomarkers Prev., June 1, 2006; 15(6): 1137 - 1141. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Le, J. T. Arnold, K. K. McFann, and M. R. Blackman DHT and testosterone, but not DHEA or E2, differentially modulate IGF-I, IGFBP-2, and IGFBP-3 in human prostatic stromal cells Am J Physiol Endocrinol Metab, May 1, 2006; 290(5): E952 - E960. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Lu, D. Katsaros, A. Wiley, I. A. Rigault de la Longrais, H. A. Risch, M. Puopolo, and H. Yu The Relationship of Insulin-Like Growth Factor-II, Insulin-Like Growth Factor Binding Protein-3, and Estrogen Receptor-{alpha} Expression to Disease Progression in Epithelial Ovarian Cancer Clin. Cancer Res., February 15, 2006; 12(4): 1208 - 1214. [Abstract] [Full Text] [PDF]
|
|
|
|
0.5) and 10% or less of controls are misdiagnosed as
positive (specificity, 