The Insulin-Like Growth Factor System and Mammographic Features in Premenopausal and Postmenopausal Women

The insulin-like growth factor (IGF) axis plays an essential role in the development of the mammary gland. IGF-I and IGF-II have potent mitogenic and antiapoptotic effects and stimulate breast cancer growth in vitro and in vivo (1, 2). High circulating IGF-I levels (as absolute concentrations or relative to its most important binding protein, IGFBP-3; ref. 3) at premenopausal ages, although not at postmenopausal ages, have been found to be associated with a subsequent increase in breast cancer risk (4-7). There is limited epidemiologic data (8, 9) on the association between circulating levels of IGF-II and subsequent breast cancer risk despite the fact that IGF-II circulates at much higher concentrations than IGF-I.

Mammographic breast density is thought to represent fibroepithelial proliferation (10). Women with extensive areas of mammographic density (≥75% of the breast area) have ∼5-fold increase in breast cancer risk compared with those with little density (<5%; ref. 11). Most studies (12-15), although not all (16), reported positive associations between circulating levels of IGF-I and/or IGF-I/IGFBP-3 molar ratio and percentage density in premenopausal women. Only one study (14), however, considered the separate associations of IGF-I and IGFBP-3 with the absolute amounts of radiologically dense (fibroepithelial) and lucent (adipose) tissues of the breast, although another (13) examined the association of these peptides with both percentage density and amount of dense tissue. The absolute amount of dense tissue in the breast may be biologically more relevant than percentage density (17) as it is likely to reflect more closely the number of epithelial cells in the breast and/or their rate of proliferation and thus their probability of suffering a malignant transformation. None of the previous studies examined mammographic features in relation to circulating levels of IGF-II.

Circulating levels of IGFs and their binding proteins are determined by a combination of genetic and environmental factors (18). Known polymorphisms in the IGF-I and IGF-II genes have not been found to be associated with serum levels of their protein products (7, 19) or with breast cancer risk (7). However, several studies have reported an association between an A/C polymorphism in the IGFBP-3 gene at the −202 locus and circulating levels of IGFBP-3 (16, 19-22), and one showed a direct association with percentage breast density (16). Mammographic density is highly heritable (23), and although the relevant genes have not yet been identified, genetic variants involved in the regulation of the IGF system would seem likely candidates.

The aim of this study is to examine the association of IGF-I, IGF-II, and IGFBP-3 serum levels with size of the radiologically dense and lucent areas of the breast. We also examined whether the previously described A/C polymorphism in the IGFBP-3 promoter region was associated with IGFBP-3 serum levels and mammographic features.

All women ages ≥34 years, resident in the island of Guernsey, United Kingdom, were invited to participate in a series of prospective studies to investigate the role of endogenous hormones in the etiology of breast cancer (8, 17, 24). A total 5,104 women (response rate = 31%) volunteered to participate in the third study, Guernsey III (GIII), in 1977 to 1985, and were subsequently invited, to participate in Guernsey IV (GIV), in 1986 to 1991 (response rate = 75%). In all, 3,679 women volunteered to participate in both studies. At entry to GIV, women completed an interviewer-administered questionnaire on lifestyle variables and had their anthropometric measurements, a nonfasting blood sample, and mammography taken. The study was approved by the relevant ethics committees, and participants gave written informed consent.

To investigate the relationship between the IGF system and mammographic features, we conducted a cross-sectional study nested within these 3,679 Guernsey women. The study was restricted to premenopausal women (those reporting they were still menstruating in their usual pattern) and natural postmenopausal women (those reporting a natural menopause whose last menstrual period was at least 1 year before entry into GIV; n = 1,168 excluded). Women were further excluded if (a) they had a history of cancer (n = 180); (b) they were on oral contraceptives or hormone replacement therapy at the time of entry into GIV (n = 115); (c) mammograms were no longer available or they had breast implants (n = 192); and (d) the interval between blood sampling and mammography was >60 days or no sample was available (n = 864). Thus, 529 premenopausal and 631 postmenopausal women were eligible, and a random sample comprising 219 premenopausal and 241 postmenopausal women was selected. IGFI, IGF-II, and IGFBP-3 measurements were obtained for 215 premenopausal and 238 postmenopausal women. Clotted blood samples adequate for IGFBP-3 genotyping were available only for a subset (139 premenopausal and 143 postmenopausal women).

Mammograms were digitized with an Astra 2400S scanner with 8-bit (0-255 grey values) output and mammographic density assessed using a specially developed computer programme (17). Readings from views of the right and left breasts and from craniocaudal and medial-lateral oblique views of the same breast are strongly correlated (r = 0.86-0.96; ref. 25); thus, only the left craniocaudal view was analyzed. For each digitized image, the observer selected interactively two threshold grey levels, one to identify the edge of the breast and a second to identify mammographic densities. The computer programme then calculated the total breast area and the dense area. Lucent area was calculated as the difference between the breast area and the dense area, and percentage breast density was calculated as the percentage of dense area within the breast area. One of the authors (G.T.M.) did all measurements without knowledge of the baseline characteristics or the IGF values of the participants. These computer-assisted measures were found to be highly reproducible in mammograms from a random sample of 102 GIV women: the within-person and between-person intraclass correlation coefficients for measurement of total breast area were both 0.99; the corresponding coefficients for percentage breast density were 0.94 (95% confidence interval, 0.91-0.98) and 0.92 (95% confidence interval, 0.89-0.94), respectively (17).

GIV serum samples were stored at −20°C until they were shipped on dry ice to the Division of Surgery, University of Bristol, United Kingdom for measurements of serum IGF-I, IGF-II, and IGFBP-3. All samples were analyzed in the same laboratory batch, technicians being blind to the baseline and mammographic characteristics of the women. Double-antibody ELISA assays were used to measure IGF-I (DSL-10-2800 Active) and IGF-II (DSL-10-2600; Diagnostic Systems Laboratories, Webster, Texas). Assays for serum IGFBP-3 were carried out using a previously validated in-house double-antibody RIA (26). Total coefficients of variations (intra-assay and inter-assay combined) in the Guernsey studies have been found to be 6.6% for IGF-I, 12.0% for IGF-II, and 3.9% for IGFBP-3 (8). The serum samples were stored for a median of 15 years (range, 13-17 years), but there was no correlation between hormone concentrations and storage time (r = 0.004 for IGF-I, r = 0.065 for IGF-II, and r = −0.026 for IGFBP-3).

Genomic DNA was extracted from 200 μL of clotted blood (following overnight digestion with 10 units streptokinase; Sigma, St. Louis, MO) using the QIAamp DNA Blood Mini kit (Qiagen, Valencia, CA). The IGFBP-3 polymorphism at position −202 relative to the CAP site was genotyped by PCR amplification using primers IGFBP-3 −392 (5′-ACCGGCTCGCCGCAGGGAGA-3′) and IGFBP-3 −119 (5′-GGGCCCGTGCTTCGCCCTGA-3′) followed by restriction digestion with HhaI. Reaction conditions are available upon request. A negative control and two IGFBP3 A/C heterozygote controls were included in each PCR. The genotype of each sample was independently assigned by two observers (100% concordance observed) and was repeated in duplicate (53% samples) or triplicate (10% samples). A proportion of samples of each genotype were sequenced for confirmation, and the genotype frequencies were determined to be in Hardy-Weinberg equilibrium.

Data for premenopausal and postmenopausal women were analyzed separately as the effect of IGFs on mammographic features may differ by menopausal status (12, 13, 15). Serum hormone levels and percentage breast density were normally distributed. A natural logarithm transformation was used to normalize the distributions of size of the dense and lucent areas. IGF-I/IGFBP-3 and IGF-II/IGFBP-3 molar ratios were estimated after converting the IGFs and IGFBP-3 values from ng/mL to nmol/L by multiplying them by factors 0.13 and 0.025, respectively. Linear regression models were fitted to assess the association of IGF levels with mammographic features, and to examine the association of IGFBP-3 genotype with IGFBP-3 serum levels and mammographic features. Separate models were fitted with quantitative exposure variables in their original continuous scale and after categorization by dividing their distributions into fourths, but as they gave similar results, only results from the latter are reported. Tests for linear trend and heterogeneity in arithmetic mean of the outcome variable (or of its logged values) across exposure categories were calculated using likelihood ratio tests (27). Ps are two sided. Statistical analyses were carried out in STATA (28).

The characteristics of the study women are summarized in Table 1. There were no marked differences between these women and those who would have been eligible if an appropriate serum sample had been available. Mean IGF-I levels were higher in premenopausal than in postmenopausal women, but there were no clear differences in IGF-II or IGFBP-3 levels by menopausal status. Size of dense area and percentage breast density were higher in premenopausal women, whereas the size of lucent area was higher in postmenopausal women. IGF-I levels were positively correlated with IGF-II levels in premenopausal and postmenopausal women (Pearson's correlation coefficient, r = 0.38, P < 0.001; r = 0.39, P < 0.001, respectively). IGFBP-3 levels were positively correlated with both IGF-I levels (r = 0.54, P < 0.001; r = 0.54, P < 0.001) and IGF-II levels (r = 0.58, P < 0.001; r = 0.65, P < 0.001) in premenopausal and postmenopausal women, respectively.

Age was associated with decreasing levels of IGF-I (P for change in mean level with 1-year increase in age = 0.01) and IGFBP-3 (P < 0.001) but not IGF-II in postmenopausal women only. After adjustment for age and time since blood collection, levels of IGF-II and IGFBP-3 but not IGF-I increased with increasing body mass index (BMI) and waist circumference (P for change in mean IGF-II with 1-unit increase in BMI < 0.001 in both premenopausal and postmenopausal women; corresponding Ps for IGFBP-3: 0.006 and <0.001, respectively). In premenopausal women, IGF-I was negatively associated with being a current smoker (P = 0.03) and a past oral contraceptive user (P = 0.03). IGFBP-3 was also negatively associated with past oral contraceptive use (P = 0.01) among premenopausal women, whereas IGF-II was negatively associated with age at first birth among postmenopausal parous women (P for change in mean IGF-II with 1-year increase in age at first birth = 0.03).

In premenopausal women, there was some evidence of an increase in the dense area with increasing fourths of IGF-I, IGF-II, and IGFBP-3, and these trends became stronger after adjustment for body adiposity and other variables (Table 2). After further adjustment of the IGF-I and IGF-II effects for IGFBP-3 levels and of the IGFBP-3 effect for IGF-I and IGF-II levels, none remained statistically significant. Lucent area was positively associated with IGF-II and IGFBP-3, but these associations were no longer significant after adjusting for BMI and waist circumference as these variables were strongly correlated with lucent area (r ∼0.66-0.71, P < 0.001 in both premenopausal and postmenopausal women). Percentage density was not associated with IGF-I, IGF-II, or IGFBP-3 in premenopausal women. In postmenopausal women, there was no evidence of an association between dense area and IGF-I, IGF-II, or IGFBP-3 before or after adjustment for body adiposity and other variables (Table 3). These peptides were, however, positively associated with lucent area and inversely associated with percentage density, but these trends were no longer significant after adjusting for body adiposity. Further adjustment for the levels of the other peptides did not affect the results. There were no associations between IGF-I/IGFBP-3 and IGF-II/IGFBP-3 molar ratios and mammographic features in premenopausal or postmenopausal women.

There were no marked baseline, mammographic, or IGF serum differences between genotyped and nongenotyped women. As observed for the whole study population, there was a positive association between IGFBP-3 serum levels and lucent area in both premenopausal and postmenopausal genotyped women (data not shown). A positive linear trend in the mean IGFBP-3 serum levels by number of A alleles at the −202 locus was observed, although in premenopausal women, this trend was attenuated after adjusting for body adiposity and other variables (Table 4). This SNP accounted for 5% (R² = 0.05) of the total variance in IGFBP-3 serum levels in premenopausal and 16% (R² = 0.16) in postmenopausal women. There were no linear associations between number of A alleles and dense area or percentage density. In premenopausal women, the number of A alleles was positively associated with lucent area, but this trend was no longer statistically significant after adjustment for body adiposity.

One of the strengths of this study is that it was possible to examine the effects of IGF-I and IGFBP-3 separately on the amounts of radiologically dense and lucent breast tissues. This study was also the first to have examined the role of serum levels of IGF-II and the second (16) to have assessed the role of IGFBP-3 genotype. The reliability of the mammographic measurements was very high (17), and both percentage density and size of the dense area were strong predictors of subsequent breast cancer risk in the Guernsey studies (17). Variations in breast density through the menstrual cycle have been reported (29), but they are far too small to have affected substantially our findings in premenopausal women. Our IGF measurements were based on a single serum sample, but this seems to be adequate to reliably measure long-term levels of these peptides (30, 31). Although the samples were stored at −20°C for a long period, the mean peptide levels observed in this study were similar to those reported by others (7) and uncorrelated with time since blood collection.

Our findings are consistent with high IGF-I serum levels being associated with modest increases in the size of the dense area in premenopausal women. IGF-I levels were found to be positively associated with percentage density in premenopausal women in all previous studies (12-15) but one (16). The direction of the IGF-I dense area association is compatible with the positive association between premenopausal IGF-I levels and breast cancer risk found in practically all prospective studies that have examined this relationship (4-7).

Our study is consistent with a possible positive association, although only of borderline statistical significance, between IGF-II levels and dense area in premenopausal women. No association between IGF-II and breast cancer risk was observed in the Guernsey studies (8), the only prospective study to have thus far examined this relationship in premenopausal women. A positive association between postmenopausal IGF-II levels and risk of breast cancer was reported in one recent prospective study (9).

Previous studies have reported either an inverse (12-15) or a null (16) association between IGFBP-3 levels and percentage density and/or size of dense area. Our observation of a significant positive relationship between IGFBP-3 and dense area in premenopausal women is, however, consistent with a positive association between premenopausal levels of this peptide and risk of breast cancer reported by most, but not all, prospective studies that have examined the IGFBP-3/breast cancer association (4-7). Some of the discrepancies in the literature regarding the role of IGFBP-3 might be related to assay methodology (32). We used a well-validated assay that is calibrated against recombinant glycosylated IGFBP-3 that dilutes in parallel with the IGFBP-3 being measured in serum samples (26).

The modest associations between IGF-I, IGF-II, and IGFBP-3 and dense area in premenopausal women observed in this study suggest a direct growth-stimulating effect of the IGF system on the fibroepithelial tissue of the breast. The individual role of each one of these three peptide is, however, difficult to establish as their levels were strongly correlated. IGFs can also be produced locally and act in an autocrine/paracrine way. Breast tissue sections from premenopausal subjects with extensive densities were found to have larger stained areas of IGF-I when compared with subjects with little density (33). Thus, it is conceivable that mammographic density may be a better surrogate for IGFs/IGFBP-3 bioavailability in the breast tissue than circulating levels of these peptides. The fact that the IGFs and IGFBP-3 associations with dense tissue, and with breast cancer risk, are restricted to premenopausal women may indicate a more important role of the IGF axis early in life and/or a synergistic effect with sex hormones (34).

The positive associations of IGF-II and IGFBP-3 levels with lucent area in premenopausal and postmenopausal women, which disappeared after adjustment for body adiposity, are likely to reflect the associations of these peptides with body adiposity, and the relation of body adiposity with the amount of lucent tissue in the breast. Similarly to our study, others, although not all (35, 36), reported positive associations between BMI and IGFBP-3 but no association between BMI and IGF-I (37-39).

This study, like others (16, 19-22), showed a positive association between the number of A alleles at a previously described functional polymorphic locus in the promoter region of the IGFBP-3 gene and IGFBP-3 serum levels in both premenopausal and postmenopausal women. Deal et al. (20) found a significantly higher promoter activity of the A allele compared with the C allele in vitro, suggesting that this IGFBP-3 polymorphism influences gene expression. The only other study to have thus far examined the role of this polymorphism on mammographic features reported positive associations of number of A alleles with both IGFBP-3 serum levels and percent density in premenopausal women only, but surprisingly, IGFBP-3 serum levels had no effect on percentage density (16). We found no association between IGFBP-3 genotype and dense area (or percent density) but the study had limited power to detect small differences (about 250 women in each genotype stratum would be required to ensure that the study had 90% power, at the 5% significance level, to detect the 10% absolute difference in mean percentage density between genotypes AA and CC reported by Lai et al.; ref. 16).

The study of hormonal determinants of mammographic features is important as these may be used as intermediate markers in studies investigating the etiology of breast cancer or testing new preventive strategies (40, 41). The cross-sectional design of the study makes it impossible to establish temporality, but the modest associations of IGF-I, IGF-II, and IGFBP-3 serum levels with the amount of mammographically dense tissue in premenopausal women are consistent with the hypothesis that the IGF system may somehow promote growth of the fibroepithelial tissue in the breast.