Insulin-like growth factor-I (IGF-I) stimulates cell proliferation and can enhance the development of tumors in different organs. Epidemiologic studies have shown that an elevated level of circulating IGF-I is associated to increased risk of breast cancer as well as other cancers. Genetic variants affecting the release or biological action of growth hormone (GH), the main stimulator of IGF-I production, may predict circulating levels of IGF-I and have an effect on cancer risk. We tested this hypothesis with a large case-control study of 807 breast cancer patients and 1,588 matched control subjects nested within the European Prospective Investigation into Cancer and Nutrition. We genotyped 22 common single nucleotide polymorphisms in 10 genes involved in GH production and action (GHRH, GHRHR, SST, SSTR1-SSTR5, POU1F1, and GH1), and in parallel, we measured serum levels of IGF-I and IGFBP-3, its major binding protein, in samples of cases and controls. SST and SSTR2 polymorphisms showed weak but statistically significant associations with breast cancer risk. SSTR5 polymorphisms were associated with IGF-I levels, whereas one polymorphism in GHRHR and one in POU1F1 were associated with IGFBP-3 levels. Our conclusion is that common genetic variation in the GH synthesis pathway, as measured by single nucleotide polymorphisms selected in the present study, is not a major determinant of IGF-I and IGFBP-3 circulating levels, and it does not play a major role in altering breast cancer risk.

There is substantial evidence that insulin-like growth factor-I (IGF-I) can enhance the development of tumors in different organs. Studies in vitro have shown that IGF-I inhibits apoptosis and stimulates cell proliferation in a variety of cell types (1). Furthermore, tumor development can be strongly enhanced in animals or organs that have been genetically or otherwise manipulated to either overexpress IGF-I or the IGF-I receptor, whereas animals made deficient in IGF-I are protected (2). In humans, several epidemiologic studies have shown increased risks of cancers of the breast (3-7), colon (8-11), prostate (12-15), and possibly other organs (16) among women and men who have comparatively elevated blood levels of IGF-I, measured either as absolute concentrations or relative to its principal plasma binding protein, IGFBP-3. The association of circulating IGF-I levels with breast cancer risk thus far has been particularly apparent among women of premenopausal or early menopausal age (3-5).

Nutrition, especially the availability of energy and amino acids, is a key determinant of circulating IGF-I levels (17, 18). Besides nutrition, however, heritability studies have shown that in Western populations a large part (40-60%) of variation in IGF-I is determined by genetic factors (19-21). Although current research to identify genetic determinants of circulating IGF-I and IGFBP-3 is intensifying (22-25), thus far, few studies have been conducted to search comprehensively for polymorphisms in genes directly or indirectly involved in regulating IGF-I synthesis and to correlate these with intersubject variations in IGF-I and IGFBP-3 levels or cancer risk.

The main endocrine stimulus of hepatic and tissue production of IGF-I and IGFBP-3 is growth hormone (GH). Therefore, examination of genetic variants, which could affect the pituitary release or biological action of GH, may be one way of predicting circulating levels of IGF-I (22). In addition to the gene encoding human GH itself (GH1), major candidate genes to be examined are those involved in controlling the pituitary synthesis and release of GH. The latter include GH releasing hormone (GHRH) and its receptor (GHRHR) as well as somatostatin (SST) and its receptors (SSTR1-SSTR5), which enhance or inhibit the synthesis and release of GH, respectively. A pituitary-specific transcription factor, called POU domain class 1 transcription factor 1 (POU1F1), is also centrally involved in regulating GH synthesis.

For each of these genes, polymorphisms that change gene expression or protein function might result in a relative increase or decrease in circulating IGF-I or IGFBP-3 levels. In several of these genes, rare genetic mutations have been identified that result in radically altered hormone levels and in growth-related diseases, such as acromegaly or familial dwarfism (26-29). However, only a few studies have shown associations between more common polymorphisms and variation of IGF-I levels compatible with normal physiology (23, 25).

To examine whether common genetic variants of GHRH, GHRHR, SST, SSTR1-SSTR5, POU1F1, and GH1 were associated with variations in circulating IGF-I and IGFBP-3 levels and possibly also with breast cancer risk, we conducted a large case-control study of 807 breast cancer patients and 1,588 matched control subjects nested within the cohorts of the European Prospective Investigation into Cancer and Nutrition (EPIC; refs. 30, 31). For the present study, an attempt was made to include all known, common polymorphisms that have the highest chance of having an effect on gene expression or function of the gene product.

The EPIC Study

The EPIC cohort consists of ∼370,000 women and 150,000 men, ages 35 to 69 years, recruited between 1992 and 1998 in 23 research centers in 10 Western European countries. The vast majority (>97%) of subjects recruited in the EPIC cohort are of European (Caucasian) origin. EPIC study subjects provided anthropometric measurements (height, weight, waist, and hip circumferences) and extensive, standardized questionnaire information about medical history, diet, physical activity, smoking, and other lifestyle factors. Women also answered questions about menstrual and reproductive history, hysterectomy, ovariectomy, and use of exogenous hormones for contraception or treatment of menopausal symptoms. In addition, ∼240,000 women and 140,000 men provided a blood sample, and plasma, serum, red cells, and a buffy coat were stored for future analyses on cancer cases and controls. Cohort members are contacted periodically to renew information on smoking, alcohol drinking, physical activity, weight, menstrual history, pregnancies, use of medications and exogenous hormones, hysterectomy, and first development of major diseases other than cancer (e.g., diabetes).

Cases of cancer occurring after recruitment into the cohort are identified through local and national cancer registries in 7 of the 10 countries and in France, Germany, and Greece by a combination of contacts with national health insurances and/or active follow-up through the study subjects or their next of kin. Follow-up on vital status, to monitor the population remaining at risk for cancer, is achieved through record linkage with mortality registries. In all EPIC study centers, cancer diagnosis is confirmed through comprehensive review of pathology reports, and checks for completeness of follow-up are made periodically. A fully detailed description of the EPIC study has been published elsewhere (30, 31).

Selection of Case and Control Subjects

Cases and controls from the present study were from 16 of the 23 EPIC recruitment centers in 7 of the 10 countries (United Kingdom, Germany, the Netherlands, France, Spain, Italy, and Greece), and most were also part of nested case-control studies on serum hormones and breast cancer risk reported in detail elsewhere22

22

R. Kaaks et al. Postmenopausal serum androgens, oestrogens and breast cancer risk: the European Prospective Investigation into Cancer and Nutrition (EPIC), in press.

,23
23

Rinaldi et al., in preparation.

(32).

Case subjects were selected among women who developed breast cancer after their recruitment into the EPIC study, and before the end of the study period, for each study center defined by the latest end-date of follow-up. Women who used any hormone replacement therapy at the time of blood donation or any exogenous hormones for contraception or medical purposes and who had previous diagnosis of cancer (except nonmelanoma skin cancer) were excluded from the study, because each of these various factors could have altered circulating hormone levels.

For each case subject with breast cancer, two control subjects were chosen at random from among cohort members alive and free of cancer (except nonmelanoma skin cancer) at the time of diagnosis of the index case. Control subjects were matched to the cases by study center where the subjects were enrolled in the cohort as well as by menopausal status (premenopausal, postmenopausal, or perimenopausal/undefined), age (±6 months) at enrollment, follow-up time, fasting status, time of the day of blood donation, and phase of the menstrual cycle for premenopausal women22 (32).

Approval for the study was given by the relevant ethical committees both at the IARC and in the EPIC recruitment centers.

Identification and Selection of Single Nucleotide Polymorphisms

We collected data on polymorphisms from publicly available databases, such as dbSNP (http://www.ncbi.nlm.nih.gov/SNP/), SNPper (http://snpper.chip.org/), and Frequency Finder (http://bluegenes.bsd.uchicago.edu/frequencyfinder/). We complemented database searches with literature review and for some genes (SST, SSTR1-SSTR5, GHRH, GHRHR, and POU1F1) with analysis of 95 subjects from the EPIC population by denaturing high-performance liquid chromatography (DHPLC; ref. 33).

To be included in the study, polymorphisms had to be located in exons (including untranslated regions), exon-intron junctions, or promoter regions of a gene of interest or otherwise should be within intronic regions that showed >80% homology between human and mouse (as reported by the University of California at Santa Cruz Genome Browser, http://genome.ucsc.edu/) and thus were likely to harbor regulatory sequences. In addition, we also included polymorphisms with documented evidence of their existence in Caucasians according either to literature data or to our own experimental analysis by DHPLC. Among all polymorphisms thus identified, we only retained those with a minor allele frequency ≥5% in Caucasians or those that result in an amino acid change and had a minor allele frequency ≥1%. Finally, we particularly favored the inclusion of all polymorphisms reported previously to be associated with cancer and/or levels of circulating hormones.

Collecting information on polymorphisms from the literature, public databases, and our own experimental analyses by DHPLC provided a list of 74 single nucleotide polymorphisms (SNP). All new SNPs identified in our laboratory by DHPLC searches have been deposited in dbSNP (http://www.ncbi.nlm.nih.gov/SNP/). By applying the selection criteria outlined above, we selected 32 SNPs for genotyping. For 2 SNPs, genotyping assays could not be designed (i.e., specialized algorithms were unable to find suitable PCR primers and/or TaqMan probes), and for 8 more SNPs, TaqMan assays were generated but provided poor genotyping results (i.e., insufficient amplification and/or insufficient separation of genotype clusters). This left 22 polymorphisms that were genotyped on the DNAs of cases and controls (Table 1). The number of SNPs typed per gene ranged from one for the small somatostatin receptors SSTR1, SSTR3, and SSTR4 to five for GHRHR.

Genotyping

Buffy coat samples for the study subjects were retrieved from the EPIC biorepository and DNAs were extracted on an Autopure instrument (Gentra Systems, Minneapolis, MN) with Puregene chemistry (Gentra Systems).

Genotyping was done by the 5′ nuclease assay (TaqMan). The order of DNAs from cases and controls was randomized on PCR plates to assure that an equal number of cases and controls could be analyzed simultaneously. TaqMan probes were synthesized by either Applied Biosystems [with minor groove binder (MGB) chemistry] or Proligo (Paris, France) [with or without locked nucleic acid (LNA) chemistry]. Sequences of primers and probes are reported in Appendix 1. The reaction mix included 10 ng genomic DNA, 5 pmol of each primer, 1 pmol of each probe, and 2.5 μL of 2× Master Mix (Applied Biosystems Foster City, CA) in a final volume of 5 μL. The thermocycling included 50 cycles with 30 seconds at 95°C followed by 60 seconds at 60°C. PCR plates were read on an ABI PRISM 7900HT instrument (Applied Biosystems). To validate genotype identification, we repeated 8% of all genotypes. Laboratory personnel was kept blinded to case-control status throughout the study.

Hormone Measurements

Measurements of IGF-I and IGFBP-3 were done in the laboratory of the Hormones and Cancer Team at IARC using ELISAs from Diagnostic System Laboratories (Webster, TX). The IGF-I assays included an acid-ethanol precipitation step to eliminate IGF-I binding proteins to avoid their interference with the IGF-I measurement. Measurements were done on never-thawed serum sample aliquots. The mean intrabatch and interbatch coefficients of variation were 6.2% and 16.2%, respectively, for IGF-I and 7.2% and 9.7%, respectively, for IGFBP-3.

Statistical Analysis

We reconstructed individuals' haplotype frequencies (i.e., estimated numbers of copies of haplotypes) using the program “tagSNPs” (http://www-rcf.usc.edu/∼stram/tagSNPs.html; refs. 34, 35). This program calculates, for each individual, the expected numbers of copies (“dosages”) of each of the haplotypes compatible with the individual's SNP genotypes. This method takes into account uncertainties in the haplotype reconstruction for individuals that are heterozygote for two or more of the SNPs within a given gene. Haplotype dosages are estimated from the individuals' SNP genotype data and from overall haplotype frequency estimates for the full study population (cases and controls combined) estimated by a maximum likelihood method. For each haplotype, the dosage values range from 0 to 2.0 (alleles), and for each individual, these dosage values add up to a total value of 2.0 across all possible haplotypes.

All association analyses, at the level of individual SNPs or gene loci, were done under different assumed modes of inheritance of effect (dominant, recessive, or codominant) associated with alleles. In the “dominant” model, circulating peptide levels or disease risks were compared between subjects carrying at least one copy of the rare allele and those who had none; in the “recessive” model, the comparison was between those who were homozygous for the rare allele and all others; in the “codominant” model, individuals' peptide levels or the logarithm of disease risk were linearly related to the number of copies of an allele (0, 1, or 2 for SNP alleles or dosages for the haplotype) carried by the individuals. For rare alleles, with a frequency <20% (i.e., a prevalence of homozygous recessive allele carriers <4.0%), only the dominant model was used.

Relationships of polymorphic gene variants with serum levels of IGF-I and IGFBP-3 were estimated by standard regression models, stratified by EPIC recruitment center, and further adjusted for age. These analyses were done both using all the study subjects and only the controls, who represent the population giving rise to the cases. Relationships of polymorphic variants with breast cancer risk [odds ratios (OR)] were estimated using conditional logistic regression models applied on the matched case-control sets. Both series of analyses were done at the level of single SNP loci as well as at the level of haplotypes (using the haplotype dosage values). Haplotype analyses were done at the level of full gene loci (i.e., including haplotypes based on all of the SNPs in that gene). In all haplotype analyses, the most common haplotype was used as the reference category.

Subgroup analyses on women with a breast cancer diagnosis either before (45% of the subjects) or after age 55 years were done, and possible heterogeneity of effect between these two age groups was tested using a χ2 test.

Eight hundred seven incident cases of breast cancer from the EPIC cohort and 1,588 matched controls were included in our study. Table 2 summarizes the baseline characteristics of cases and controls. The mean age of study subjects was 55 years. For cases, the mean age at diagnosis was 57 years. Based on the questionnaire data, 32% of the subjects were premenopausal at blood donation, 10% were perimenopausal or of unknown menopausal status, and 58% were postmenopausal. Cases had a significantly lower number of full-term pregnancies than controls (mean, 2.35 versus 2.47; P = 0.02) and were significantly older at first full-term pregnancy (26 versus 25.5 years in controls; P = 0.02). Age at menarche did not differ between cases and controls, nor did body mass index or height. Serum levels of IGF-I adjusted for age and center were not significantly different between cases and controls (mean, 248.7 versus 244.4 ng/mL; P = 0.15), nor for the subgroups of subjects with cancer diagnosis before or at age 55 years (272.0 versus 270.7; P = 0.79) or after age 55 years (224.0 versus 217.5; P = 0.08). Case subjects did show higher mean levels of serum IGFBP-3 than controls (mean, 3,422 versus 3,361 ng/mL; P = 0.04). The latter difference, however, was due mostly to the subgroup with cancer diagnosis after age 55 years (3,473 versus 3,378; P = 0.02) and was not clearly visible among the younger women (3,190 versus 3,173; P = 0.66). IGF-I and IGFBP-3 levels were significantly correlated (Pearson partial correlation coefficient on log-transformed data = 0.43; P < 0.0001). Details on the relationships of IGF-I and IGFBP-3 with breast cancer risk, with an extended series of 1,081 breast cancer cases and 2,098 control subjects, are reported elsewhere.23

Genotyping call rates ranged between 95.36% and 99.53%. The distributions of genotypes of all polymorphisms were in agreement with Hardy-Weinberg equilibrium (calculated in the control group). Repeated quality control genotypes (8% of the total) showed >99% concordance for all assays.

Results of associations between individual SNPs and cancer risk and circulating hormone levels are reported in Table 3. Table 4A to E reports results of analyses of haplotypes of genes for which two or more polymorphisms have been typed.

Serum concentrations of IGF-I showed a nominally significant association only with one SNP (rs642249; P = 0.002) in SSTR5 but not with any polymorphisms in the other genes studied. Haplotype analysis did not add anything to this finding, as there was a significant association with the only SSTR5 haplotype (hCCCA), which includes the A allele of rs642249. When we stratified the data by age at diagnosis (before or after age 55 years), the association was observed only in women ages <55 years, but the heterogeneity test was not significant (data not shown). Significant associations with reduced levels of IGF-I were also observed for two haplotypes of GH1 (i.e., hTAAT and hGGAT), although no single SNP of this gene was associated with IGF-I level. We observed the same associations when we restricted analysis to the controls.

Serum IGFBP-3 concentrations were significantly increased among carriers of the rs300982 C allele in POU1F1 (P = 0.01 and 0.01 for codominant and dominant models, respectively) both when we analyzed all the study subjects together and when we analyzed only the controls. A stratified analysis showed that the association was present only in the older women, although a test for heterogeneity was not significant.

In addition, analyses restricted to the controls identified also a weak (P = 0.04 for the dominant model) association with SNP rs2228487 of SSTR1.

Breast cancer risk showed statistically significant associations (P < 0.05) with polymorphic variants in the SST, SSTR2, and GH1 genes.

For the SST gene, carriers of two different SNPs showed an increase in breast cancer risk, with relative risks of ∼1.3 for both rs4988513 C and P0689 C alleles. Reflecting these two associations, the hCC haplotype of SST also showed an effect on risk [OR, 1.27; 95% CI (95% CI),1.02-1.59 for the dominant model]. In analyses stratified by age at diagnosis, a statistically significant increase in risk was observed only in the higher age group (OR, 1.53; 95% CI, 1.17-2.01 for carriers of the rs4988513 C allele; OR, 1.41; 95% CI, 1.07-1.88 for carriers of the P0689 C allele) but not among women with a breast cancer diagnosis at age <55 years. However, interaction tests showed no statistically significant heterogeneity of effect between the two age groups.

For the SSTR2 gene, breast cancer risk was decreased among homozygous carriers of the C allele of SNP rs1466113 (OR, 0.74; 95% CI, 0.57-0.96). This reduction in risk was mirrored by an increased risk associated with the haplotypes bearing the other allele (hGG: OR, 1.24; 95% CI, 1.03-1.51; hAG: OR, 1.19; 95% CI, 0.99-1.45). Heterozygosity at the other SSTR2 polymorphism we typed (rs998571) was associated with a nearly significant increase in breast cancer risk (OR, 1.20; 95% CI, 0.99-1.45).

For the GH1 gene, only subjects who were heterozygous for the rs6171 allele showed an association with reduced cancer risk (OR, 0.77; 95% CI, 0.63-0.94), which was compatible, however, with a dominant effect of the G allele toward a reduction in risk (P = 0.03).

We have done a large-scale association study, nested in the EPIC cohort, to assess the role of genetic variation of genes involved in the GH synthesis pathway on risk of breast cancer and on circulating levels of IGF-I and IGFBP-3. The present study was conducted on a large subset of the breast cancer cases and matched control subjects that were included in a parallel study on breast cancer risk in relation to IGF-I and IGFBP-323 as well to circulating sex steroids22 (32). To our knowledge, this is the first study to look in a comprehensive way at breast cancer risk and genetic variation in the GH synthesis pathway, including a total of 10 different genes, and the first study to simultaneously analyze associations of genetic variants with circulating levels of IGF-I and IGFBP-3 in the European population.

Our objective was to include into our analyses all SNPs that would have a minimum allele frequency of 5% or otherwise a high chance of having an effect on gene expression or function of the gene product (e.g., known coding variants). We therefore did extensive searches through the literature and public databases. Although the DHPLC technique approach may be somewhat less sensitive than a systematic resequencing for the identification of new SNPs, it is a quite reliable method for SNP detection (reviewed in ref. 33). Although systematic resequencing would have probably led to the identification of further polymorphisms, this approach did not fall within the financial scope of our project. We believe, however, that most of the additional SNPs that could have been identified by such more stringent approach probably will have an allele frequency <5% and that resequencing would have led to the identification of only very few additional common polymorphisms, with higher allele frequencies. Overall, therefore, we are confident that we have included in our study most of the common variants existing in Caucasians in the 10 candidate genes examined in this study. Because the SNPs in our study were not specifically selected according to a haplotype-tagging approach, it has to be noted that our haplotype analysis could miss associations with unknown genetic variants of the candidate genes that are not in linkage disequilibrium (LD) with the SNPs we selected.

Although we have tried to have a fair representation of all common variants, we took great care in retaining for our study those polymorphisms for which evidence of experimental validation was available. This was particularly relevant for SNPs of the GH1 gene, given that it belongs to a cluster of five genes that all share very high degree of homology even for intronic and promoter regions (36). Although many polymorphisms have been reported in the GH1 gene, several of these may have been actually artifacts due to nonspecific amplification of target sequences in one of the other GH homologues. In this regard, it is important to note that none of polymorphisms we have typed in the GH1 gene showed any significant departure from Hardy-Weinberg equilibrium, which suggests that our genotyping assays for GH1 gene polymorphisms have been properly designed.

Our study population included women from 7 of the 10 countries participating in the EPIC project ranging from southern Europe (Greece, Italy, and Spain) to the United Kingdom. Over 97% of EPIC subjects are estimated to be of Caucasian origin. Nevertheless, there is substantial regional difference in breast cancer incidence rates across Europe (most likely due to differences in lifestyle). Thus, spurious associations of risk with allelic variants could be found if allelic frequencies varied substantially between regions. This potential bias was avoided, however, by matching the control subjects to the breast cancer cases by EPIC recruitment center and by performing a conditional logistic regression analysis. Regression models relating IGF-I or IGFBP-3 to polymorphic variants were also systematically adjusted for the factor “recruitment center.”

Serum levels of IGF-I and IGFBP-3 showed statistically significant associations with variants (SNPs and haplotypes) in several of the candidate genes studied. However, in spite of the large size of this cross-sectional study component, most of these associations were not highly statistically significant (Ps = 0.01-0.05). The only exception was the association between a synonymous polymorphism in exon 1 of SSTR5 and IGF-I level supported by a P of 0.002. This novel finding will have to be confirmed by further epidemiologic and/or functional studies. Furthermore, for none of the SNPs found to be associated with serum peptide levels was there any previous evidence of a similar association or any experimental evidence for a possible direct, functional role. This makes it difficult to assess whether the associations observed in our study represent a true effect on serum peptide levels (directly or through LD with other, functional polymorphisms) or whether they were merely chance findings. Additional large association studies will be needed to confirm our findings.

With regard to breast cancer risk, the only relatively consistent pattern of associations was with variants of the SST and SSTR2 genes. It is possible that associations of breast cancer risk with polymorphisms in the SST and SSTR genes reflect autocrine or paracrine mechanisms of cellular proliferation active at the level of the breast epithelium. Both somatostatin (37) and its receptors (38) have been found to be produced by breast cancer cells, and all five somatostatin receptors have been found in human breast tumors, with subtype 2 occurring more frequently (39). The presence of somatostatin receptors in human breast cancers has been correlated with well-differentiated tumors and more favorable prognosis (38), and treatment with somatostatin or its synthetic analogues has been shown to have antiproliferative effects on breast tumors, both in vitro and in vivo (40), and to result in a positive tumor response in >40% of patients (41).

In view of the above observations, it is possible that polymorphisms, which affect expression of the somatostatin gene SST and/or its receptor SSTR2, could modulate the antiproliferative effect that somatostatin exerts on breast tumors. The SST SNPs that we found to be associated with cancer risk are located in intron 1 of this gene, and nothing is known of their possible function. The association we report either may reflect a direct functional effect of the polymorphisms studied or may be due to LD with unknown functional polymorphisms. On the other hand, the SSTR2 SNPs we have typed are located in the promoter, respectively, at positions −57 and −83 upstream of the start of transcription. Interestingly, the SSTR2 haplotype (hGG) that shows an increased risk of breast cancer has been also reported to be associated with a 60% to 70% reduction of SSTR2 transcription in pancreatic cancer cells by use of site-directed mutagenesis and a luciferase reporter gene assay (42). It has to be noted that, at the individual SNP level, it was the polymorphism at position −83 that was found to be responsible for this decrease (42). We found a borderline, nonsignificant association with the SNP at the −83 position and a significant association with the SNP at −57 position. In our sample, there is a complete but not perfect LD between these two polymorphisms (D′ = 1, r2 = 0.36), reflecting the fact that their frequencies are different (minor allele frequencies of 32% and 44%, respectively). It is difficult therefore to say whether the associations we observe and the previously reported functional role for one of the SSTR2 promoter SNPs are in relation or not.

Only one SNP included in our work has been studied previously in relation to breast cancer or circulating levels of IGF-I and IGFBP-3. SNP rs2665802, located in intron 4 of GH1, has been found to be associated with level of IGF-I and risk of colorectal neoplasia (43). Another study found an association of this polymorphism with secretion of GH and IGF-I and with stature in a group of Japanese prepubertal short children (24). The same polymorphism did not show any association with breast cancer risk, however, in a large case-control study done in a Chinese population, where IGF-I levels were not measured (44). Likewise, we did not observe any association of this polymorphism with cancer risk or hormone level in our study. Nevertheless, we have found two GH1 haplotypes that showed a weak association with reduced IGF-I level. This leaves the possibility that the previously reported associations reflected LD between SNP rs2665802 and other polymorphisms in GH1 or possibly in neighboring genes.

Previous prospective cohort studies have shown increased prediagnostic IGF-I levels among women who developed breast cancer, especially when the cancer was diagnosed at a relatively young (premenopausal or early menopausal) age. Contrary to these previous findings, however, our data from the EPIC cohort, with an extended series of breast cancer cases (n = 1,081) and control subjects (n = 2,098), showed a weak, direct association of serum IGF-I with breast cancer risk only among women of postmenopausal age but not among the younger women (although mean levels did not differ between the two groups).23 In the present study, we also did analyses of genetic variants in relation to breast cancer risk stratified by age at breast cancer diagnosis (age <55 and >55 years). Our study was not large enough, however, to allow for statistically powerful tests for differences in associations between older and younger women, especially in relation to the rarer polymorphisms.

This large study, for the first time, investigated the role of genetic variation across 10 different genes belonging to the same pathway. This obviously entailed a large number of statistical tests, which may have led to several spurious associations due to chance. One approach to account for the multiple comparisons is to use Bonferroni's method, applying a more stringent criterion for statistical significance, at the level of each gene studied. This method is conservative, however, as it is difficult to account for dependence between statistical tests due to LD between SNPs. An alternative is to apply a Bayesian approach to calculate false-positive response probabilities (FPRP; ref. 45). We have computed FPRPs for the nominally significant associations we have observed between SNPs and breast cancer risk. Use of a prior probability of 0.1 resulted in noteworthy FPRPs for the association with breast cancer risk of polymorphisms rs6171 in GH1 (FPRP = 0.16) and rs4988513 in SST (FPRP = 0.16). Wacholder et al. suggest that for a large study like ours FPRP < 0.2 might be an appropriate threshold for noteworthiness (45). When using a prior probability of true association of ≤0.01, which is more likely to be correct, we obtained high FPRP values, all >0.67.

In conclusion, the associations we report here do not have a strong statistical support making replication the key to confirm or dismiss these results. Given that associations with individual genetic variants seem to be of a relatively small magnitude, even larger studies will be needed to confirm our findings and to allow for association studies on rarer polymorphic variants as well.

Grant support: U.S. Army Medical Research and Materiel Command grant DAMD17-01-0275 and Association pour la Recherche sur le Cancer grants 7610, 138, and 7478 and IARC Special Training Awards (J.D. McKay, R.J. Cleveland, and M. Llewellyn). The EPIC study was funded by “Europe Against Cancer” Programme of the European Commission (SANCO); Ligue contre le Cancer (France); Société 3M (France); Mutuelle Générale de l'Education Nationale; Institut National de la Sante et de la Recherche Medicale; German Cancer Aid; German Cancer Research Center; German Federal Ministry of Education and Research; Danish Cancer Society; Health Research Fund of the Spanish Ministry of Health; the participating regional governments and institutions of Spain; Cancer Research UK; Medical Research Council, United Kingdom; the Stroke Association, United Kingdom; British Heart Foundation; Department of Health, United Kingdom; Food Standards Agency, United Kingdom; the Wellcome Trust, United Kingdom; Greek Ministry of Health; Greek Ministry of Education; Italian Association for Research on Cancer; Italian National Research Council; Dutch Ministry of Public Health, Welfare and Sports; Dutch Ministry of Health; Dutch Prevention Funds; LK Research Funds; Dutch Zorg Onderzoek Nederland; World Cancer Research Fund; Swedish Cancer Society; Swedish Scientific Council; Regional Government of Skane, Sweden; Norwegian Cancer Society; and ISC III Network RCESP (C03/09), Spain.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Khandwala HM, McCutcheon IE, Flyvbjerg A, Friend KE. The effects of insulin-like growth factors on tumorigenesis and neoplastic growth.
Endocr Rev
2000
;
21
:
215
–44.
2
Hadsell DL, Bonnette SG. IGF and insulin action in the mammary gland: lessons from transgenic and knockout models.
J Mammary Gland Biol Neoplasia
2000
;
5
:
19
–30.
3
Hankinson SE, Willett WC, Colditz GA, et al. Circulating concentrations of insulin-like growth factor-I and risk of breast cancer.
Lancet
1998
;
351
:
1393
–6.
4
Toniolo P, Bruning PF, Akhmedkhanov A, et al. Serum insulin-like growth factor-I and breast cancer.
Int J Cancer
2000
;
88
:
828
–32.
5
Muti P, Quattrin T, Grant BJ, et al. Fasting glucose is a risk factor for breast cancer: a prospective study.
Cancer Epidemiol Biomarkers Prev
2002
;
11
:
1361
–8.
6
Yu H, Jin F, Shu XO, et al. Insulin-like growth factors and breast cancer risk in Chinese women.
Cancer Epidemiol Biomarkers Prev
2002
;
11
:
705
–12.
7
Pollak MN, Schernhammer ES, Hankinson SE. Insulin-like growth factors and neoplasia.
Nat Rev Cancer
2004
;
4
:
505
–18.
8
Ma J, Pollak MN, Giovannucci E, et al. Prospective study of colorectal cancer risk in men and plasma levels of insulin-like growth factor (IGF)-I and IGF-binding protein-3.
J Natl Cancer Inst
1999
;
91
:
620
–5.
9
Giovannucci E, Pollak MN, Platz EA, et al. A prospective study of plasma insulin-like growth factor-1 and binding protein-3 and risk of colorectal neoplasia in women.
Cancer Epidemiol Biomarkers Prev
2000
;
9
:
345
–9.
10
Kaaks R, Toniolo P, Akhmedkhanov A, et al. Serum C-peptide, insulin-like growth factor (IGF)-I, IGF-binding proteins, and colorectal cancer risk in women.
J Natl Cancer Inst
2000
;
92
:
1592
–600.
11
Palmqvist R, Hallmans G, Rinaldi S, et al. Plasma insulin-like growth factor 1, insulin-like growth factor binding protein 3, and risk of colorectal cancer: a prospective study in northern Sweden.
Gut
2002
;
50
:
642
–6.
12
Mantzoros CS, Tzonou A, Signorello LB, Stampfer M, Trichopoulos D, Adami HO. Insulin-like growth factor 1 in relation to prostate cancer and benign prostatic hyperplasia.
Br J Cancer
1997
;
76
:
1115
–8.
13
Chan JM, Stampfer MJ, Giovannucci E, et al. Plasma insulin-like growth factor-I and prostate cancer risk: a prospective study.
Science
1998
;
279
:
563
–6.
14
Stattin P, Bylund A, Rinaldi S, et al. Plasma insulin-like growth factor-I, insulin-like growth factor-binding proteins, and prostate cancer risk: a prospective study.
J Natl Cancer Inst
2000
;
92
:
1910
–7.
15
Wolk A, Mantzoros CS, Andersson SO, et al. Insulin-like growth factor 1 and prostate cancer risk: a population-based, case-control study.
J Natl Cancer Inst
1998
;
90
:
911
–5.
16
Petridou E, Koukoulomatis P, Alexe DM, Voulgaris Z, Spanos E, Trichopoulos D. Endometrial cancer and the IGF system: a case-control study in Greece.
Oncology
2003
;
64
:
341
–5.
17
Thissen JP, Ketelslegers JM, Underwood LE. Nutritional regulation of the insulin-like growth factors.
Endocr Rev
1994
;
15
:
80
–101.
18
Kaaks R, Lukanova A. Energy balance and cancer: the role of insulin and insulin-like growth factor-I.
Proc Nutr Soc
2001
;
60
:
91
–106.
19
Hall K, Hilding A, Thoren M. Determinants of circulating insulin-like growth factor-I.
J Endocrinol Invest
1999
;
22
Suppl:
48
–57.
20
Hong Y, Brismar K, Hall K, Pedersen NL, de Faire U. Associations between insulin-like growth factor-I (IGF-I), IGF-binding protein-1, insulin and other metabolic measures after controlling for genetic influences: results from middle-aged and elderly monozygotic twins.
J Endocrinol
1997
;
153
:
251
–7.
21
Harrela M, Koistinen H, Kaprio J, et al. Genetic and environmental components of interindividual variation in circulating levels of IGF-I, IGF-II, IGFBP-1, and IGFBP-3.
J Clin Invest
1996
;
98
:
2612
–5.
22
Lopez-Bermejo A, Buckway CK, Rosenfeld RG. Genetic defects of the growth hormone-insulin-like growth factor axis.
Trends Endocrinol Metab
2000
;
11
:
39
–49.
23
Jernstrom H, Deal C, Wilkin F, et al. Genetic and nongenetic factors associated with variation of plasma levels of insulin-like growth factor-I and insulin-like growth factor-binding protein-3 in healthy premenopausal women.
Cancer Epidemiol Biomarkers Prev
2001
;
10
:
377
–84.
24
Hasegawa Y, Fujii K, Yamada M, et al. Identification of novel human GH-1 gene polymorphisms that are associated with growth hormone secretion and height.
J Clin Endocrinol Metab
2000
;
85
:
1290
–5.
25
Vaessen N, Heutink P, Janssen JA, et al. A polymorphism in the gene for IGF-I: functional properties and risk for type 2 diabetes and myocardial infarction.
Diabetes
2001
;
50
:
637
–42.
26
Pernasetti F, Milner RD, al Ashwal AA, et al. Pro239Ser: a novel recessive mutation of the Pit-1 gene in seven Middle Eastern children with growth hormone, prolactin, and thyrotropin deficiency.
J Clin Endocrinol Metab
1998
;
83
:
2079
–83.
27
Ballare E, Persani L, Lania AG, et al. Mutation of somatostatin receptor type 5 in an acromegalic patient resistant to somatostatin analog treatment.
J Clin Endocrinol Metab
2001
;
86
:
3809
–14.
28
Salvatori R, Fan X, Phillips JA III, et al. Three new mutations in the gene for the growth hormone (GH)-releasing hormone receptor in familial isolated GH deficiency type Ib.
J Clin Endocrinol Metab
2001
;
86
:
273
–9.
29
Wajnrajch MP, Gertner JM, Sokoloff AS, et al. Haplotype analysis of the growth hormone releasing hormone receptor locus in three apparently unrelated kindreds from the Indian subcontinent with the identical mutation in the GHRH receptor.
Am J Med Genet
2003
;
120A
:
77
–83.
30
Riboli E, Hunt KJ, Slimani N, et al. European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection.
Public Health Nutr
2002
;
5
:
1113
–24.
31
Bingham S, Riboli E. Diet and cancer—the European Prospective Investigation into Cancer and Nutrition.
Nat Rev Cancer
2004
;
4
:
206
–15.
32
Kaaks R, Berrino F, Key T, et al. Serum sex steroids in premenopausal women and breast cancer risk within the European Prospective Investigation into Cancer and Nutrition (EPIC).
J Natl Cancer Inst
2005
;
97
:
755
–65.
33
Xiao W, Oefner PJ. Denaturing high-performance liquid chromatography: a review.
Hum Mutat
2001
;
17
:
439
–74.
34
Stram DO, Leigh PC, Bretsky P, et al. Modeling and E-M estimation of haplotype-specific relative risks from genotype data for a case-control study of unrelated individuals.
Hum Hered
2003
;
55
:
179
–90.
35
Stram DO, Haiman CA, Hirschhorn JN, et al. Choosing haplotype-tagging SNPS based on unphased genotype data using a preliminary sample of unrelated subjects with an example from the Multiethnic Cohort Study.
Hum Hered
2003
;
55
:
27
–36.
36
Chen EY, Liao YC, Smith DH, Barrera-Saldana HA, Gelinas RE, Seeburg PH. The human growth hormone locus: nucleotide sequence, biology, and evolution.
Genomics
1989
;
4
:
479
–97.
37
Nelson J, Cremin M, Murphy RF. Synthesis of somatostatin by breast cancer cells and their inhibition by exogenous somatostatin and sandostatin.
Br J Cancer
1989
;
59
:
739
–42.
38
Bootsma AH, van Eijck C, Schouten KK, et al. Somatostatin receptor-positive primary breast tumors: genetic, patient and tumor characteristics.
Int J Cancer
1993
;
54
:
357
–62.
39
Xu Y, Song J, Berelowitz M, Bruno JF. Estrogen regulates somatostatin receptor subtype 2 messenger ribonucleic acid expression in human breast cancer cells.
Endocrinology
1996
;
137
:
5634
–40.
40
Mascardo RN, Sherline P. Somatostatin inhibits rapid centrosomal separation and cell proliferation induced by epidermal growth factor.
Endocrinology
1982
;
111
:
1394
–6.
41
Dolan JT, Miltenburg DM, Granchi TS, Miller CC III, Brunicardi FC. Treatment of metastatic breast cancer with somatostatin analogues—a meta-analysis.
Ann Surg Oncol
2001
;
8
:
227
–33.
42
Torrisani J, Bouisson M, Puente E, et al. Transcription of SST2 somatostatin receptor gene in human pancreatic cancer cells is altered by single nucleotide promoter polymorphism.
Gastroenterology
2001
;
120
:
200
–9.
43
Le Marchand L, Donlon T, Seifried A, Kaaks R, Rinaldi S, Wilkens LR. Association of a common polymorphism in the human GH1 gene with colorectal neoplasia.
J Natl Cancer Inst
2002
;
94
:
454
–60.
44
Ren Z, Cai Q, Shu XO, et al. Genetic polymorphisms in the human growth hormone-1 gene (GH1) and the risk of breast carcinoma.
Cancer
2004
;
101
:
251
–7.
45
Wacholder S, Chanock S, Garcia-Closas M, El Ghormli L, Rothman N. Assessing the probability that a positive report is false: an approach for molecular epidemiology studies.
J Natl Cancer Inst
2004
;
96
:
434
–42.
46
Naylor SL, Sakaguchi AY, Shen LP, Bell GI, Rutter WJ, Shows TB. Polymorphic human somatostatin gene is located on chromosome 3.
Proc Natl Acad Sci U S A
1983
;
80
:
2686
–9.