© The Author 2006. Published by Oxford University Press.
|
|
ARTICLE |
Common Genetic Variation in IGF1 and Prostate Cancer Risk in the Multiethnic Cohort
Affiliations of authors: Department of Preventive Medicine, Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, Los Angeles, CA (IC, DOS, MP, BEH); Program in Medical and Population Genetics, Broad Institute of Harvard and Massachusetts Institute of Technology, Cambridge, MA (KLP, JH, DA, MLF); Department of Genetics (KLP, JH, DA, MLF), Medicine (KLP, DA, MLP), and Pediatrics (JH), Harvard Medical School, Boston, MA; Department of Molecular Biology (KLP, DA, MLP), Diabetes Unit (DA), and Hematology-Oncology (MLP), Massachusetts General Hospital, MA; Cancer Etiology Program, Cancer Research Center of Hawaii, University of Hawaii, Honolulu, HI (LLM, LNK); Division of Genetics and Endocrinology, Children's Hospital and Department of Pediatrics, Boston, MA (JH)
Correspondence to: Matthew Freedman, MD, Program in Medical and Population Genetics, Broad Institute of Harvard and MIT, Cambridge, MA 02139 (e-mail: freedman{at}broad.mit.edu).
 |
ABSTRACT |
|---|
 TopNotes Abstract IntroductionPatients and methodsResultsDiscussionReferences |
|---|
Background: Insulin-like growth factor I (IGF-I) appears to play a role in prostate development and carcinogenesis. We investigated whether genetic variation at the IGF1 locus is associated with prostate cancer risk. Methods: We sequenced IGF1 exons in germline DNA from 95 men with advanced prostate cancer to identify missense variants. IGF1 linkage disequilibrium patterns and common haplotypes were characterized by genotyping 64 single-nucleotide polymorphisms (SNPs) spanning 156 kilobases in 349 control subjects. Associations between IGF1 haplotypes and genotypes were investigated among 2320 patients with prostate cancer and 2290 control subjects from the Multiethnic Cohort. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by unconditional logistic regression to determine the association between prostate cancer and IGF1 haplotypes and genotypes. We used permutation testing to correct for multiple hypothesis testing. All statistical tests were two-sided. Results: No IGF1 missense variants were observed. We identified four blocks of strong linkage disequilibrium and selected a subset of 29 tagging SNPs that could accurately predict both the common IGF1 haplotypes and the remaining SNPs. Haplotype analysis revealed nominally statistically significant associations with prostate cancer risk in each of the four haplotype blocks: haplotype 1B (OR = 1.21, 95% CI = 1.04 to 1.40), haplotype 2C (OR = 1.24, 95% CI = 1.06 to 1.44), haplotype 3C (OR = 1.25, 95% CI = 1.03 to 1.50), and haplotype 4D (OR = 1.19, 95% CI = 1.02 to 1.39). Two SNPs—rs7978742 (Ptrend = .002) and rs7965399 (Ptrend = .002)—were perfectly correlated (correlation coefficient = 1.0) with one another and also associated with prostate cancer risk. These two SNPs were strong proxies for haplotypes 1B, 2C, 3C, and 4D and could account for the haplotype findings. Permutation testing revealed that a similarly strong result would be observed by chance only 5.6% of the time. Conclusion: Inherited variation in IGF1 may play a role in the risk of prostate cancer.
|
INTRODUCTION |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
Insulin-like growth factor I (IGF-I) stimulates cellular proliferation and inhibits apoptosis (1). Several lines of evidence indicate the importance of IGF-I to prostate development: IGF-I stimulates mitogenic and antiapoptotic activities of prostate epithelial cells in vitro (2–4) and is important in both normal prostate development and tumorigenesis in vivo (5–7). Prospective studies report that men with the highest levels of circulating IGF-I have a statistically significant increased risk of prostate cancer compared with men with the lowest levels of IGF-I (meta-analysis odds ratio [OR] = 2.43, 95% confidence interval [CI] = 1.11 to 5.32) (8). In the largest study to date, which evaluated 530 patients with prostate cancer in an extension of the initial report of the Physicians' Health Study, men in the highest quartile of IGF-I level had a statistically significant fivefold increased risk of being diagnosed with advanced-stage prostate cancer than men in the lowest quartile (9).
Previous studies that have investigated the role of genetic variation in IGF1 in relation to prostate cancer risk have focused solely on a (CA)n repeat sequence that is located approximately 1 kilobase (kb) upstream from the IGF1 transcription start site (10–14). Results of these studies have been inconsistent, with two studies reporting a positive association between homozygosity for the (CA)19 repeat and prostate cancer risk (10,11), one study reporting an inverse association (12), and two studies reporting no association (13,14).
To date, no study has systematically examined the genetic diversity of IGF1 in relation to prostate cancer risk. To more comprehensively evaluate this relationship, we surveyed the genetic variation of both coding and noncoding regions of IGF1 and conducted a large nested case–control study of 2320 case patients and 2290 control subjects from the Multiethnic Cohort Study.
|
PATIENTS AND METHODS |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
The Multiethnic Cohort Study is a large population-based cohort study of more than 215 000 men and women from Hawaii and Los Angeles. The cohort is composed predominantly of individuals from the following five racial or ethnic groups: African Americans, Native Hawaiians, Japanese, Latinos, and whites. Participants between the ages of 45 and 75 years were recruited from March 1993 through May 1996. Participants completed a 26-page self-administered questionnaire that included information regarding medical history, family history of cancer, diet, dietary supplements and medication use, and physical activity. Informed consent was obtained from all participants. Further details about this cohort are provided elsewhere (15).
Incident prostate cancers in participants in the Multiethnic Cohort Study were identified (up to April 1, 2002) by cohort linkage to population-based Surveillance, Epidemiology and End Results (SEER) cancer registries covering Hawaii and California. Information on stage of disease at the time of diagnosis was also collected from the cancer registries. We defined nonadvanced and advanced disease based on tumor stage and differentiation (Gleason grade). Tumors that were confined to the prostate and had a Gleason grade of less than 8 were defined as nonadvanced disease. Regional and metastatic tumors or localized tumors with a Gleason grade of 8 or more were defined as advanced disease.
Control subjects were men without prostate cancer before entry into the cohort and without a prostate cancer diagnosis up to April 1, 2002, who were randomly selected from the random control pool of Multiethnic Cohort Study participants and who provided blood specimens for genetic analyses. A total of 4177 men were contacted to participate as control subjects. The participation rate for blood collection was 72% for case patients and 69% for control subjects. Control subjects were frequency matched to case patients by age and ethnicity. This case–control study consisted of 2320 case patients with prostate cancer and 2290 control subjects. This study was approved by the institutional review boards at the University of Hawaii and the University of Southern California.
Sequencing
To identify any missense single-nucleotide polymorphisms (SNPs) in coding regions that were not in public or private databases, we attempted to sequence the four IGF1 exons in DNA from 95 case patients with advanced prostate cancer (19 per racial or ethnic group). After three sequencing attempts, exon 2 did not meet our sequencing criteria of more than 80% of the samples with phred scores (i.e., base calling scores) of 20 or more (99% accuracy of base call) for at least 80% of the target bases. Consequently, we did not sequence exon 2. Sequencing was performed by conventional dye-primer sequencing on ABI 3700 sequencers (Applied Biosystems, Foster City, CA). The PolyPhred program was used to identify polymorphisms with manual review by at least two observers (16), and all putative coding variants were validated by SNP genotyping. Further details of sequencing methods are described elsewhere (17).
SNP Selection and Genotyping for Genetic Characterization
To characterize patterns of linkage disequilibrium across the IGF1 gene, we genotyped SNPs that spanned 156 kb across the IGF1 locus (from 23.4 kb upstream of exon 1 to 47.8 kb downstream of the transcribed region). We selected SNPs from the public (http://www.ncbi.nlm.nih.gov/SNP/) and private (http://www.celera.com/) databases. Our goal was to achieve a SNP density of one SNP for every 2–5 kb across the locus. We preferentially selected SNPs located in coding and untranslated regions, as well as areas of homology to the mouse sequence (80% identity in 200-base-pair sequence; http://www.dcode.org/). No known functional polymorphisms have been reported for any IGF1 SNP. For genetic characterization, 154 SNPs were genotyped in the following multiethnic panel of 349 individuals with no history of cancer: 70 African Americans, 69 Hawaiians, 70 Japanese, 70 Latinos, and 70 whites (18). SNP genotyping was performed by the Sequenom mass spectrometry platform (Sequenom Inc., San Diego, CA). Of the 154 SNPs genotyped, 53 were identified as monomorphic and 37 had poor genotyping results (i.e., genotyped 
75% of samples or out of Hardy-Weinberg equilibrium [one-sided P<.01] in more than one ethnic group); these SNPs were eliminated from further analysis. The remaining 64 SNPs were used for genetic characterization and had an average density of one SNP for every 2.4 kb (Table 1).
|
Haplotype Block Determination and Tagging SNP Selection
The D' statistic was used to determine the pairwise linkage disequilibrium between the 64 SNPs (19). Regions of strong linkage disequilibrium (i.e., haplotype blocks) were defined by the methods of Gabriel et al. (20). To ensure thorough coverage across each haplotype block, our goal was to genotype a minimum of six SNPs with minor allele frequencies of 10% or more within each block with a minimum distance between neighboring blocks of less than 10 kb. Each ethnic group was evaluated for the extent of linkage disequilibrium that met these block criteria. Hawaiians, Japanese, Latinos, and whites shared similar linkage disequilibrium patterns. We combined these populations to assess the linkage disequilibrium structure of the locus. African Americans, as expected, displayed less linkage disequilibrium given their longer ancestral history than other groups. In an effort to characterize variation in the African American population more fully, we genotyped an additional 12 SNPs in regions where there were insufficient numbers of SNPs to meet our criteria of strong linkage disequilibrium. This additional set consisted of all publicly available SNPs in the database dbSNP as of April 2004 (http://www.ncbi.nlm.nih.gov/SNP/). Nevertheless, there was still an insufficient number of SNPs to completely fulfill our minimum six-SNP criterion for this ethnic group, indicating that more alleles may be observed with further genotyping efforts.
Genotype data for each ethnic group in the multiethnic panel were used to estimate haplotype frequencies within blocks by use of the expectation-maximization (E-M) algorithm (21). The squared correlation coefficients (
) between the true haplotypes (h) and their estimates from the E-M algorithm were calculated as described by Stram et al. (22). For each ethnic group, we then selected the minimum set of tagging SNPs, a set of informative SNPs, within each block to assure an of at least .7 for all haplotypes with an estimated frequency of at least 5% (22).
Genotyping in the Case–Control Study
The tagging SNPs were genotyped in the prostate case–control study by use of the 5' nuclease TaqMan allelic discrimination assay (Applied Biosystems). All assays were performed by investigators who were blinded to the case–control status of samples. For quality control, 5% of replicate samples were included. The concordance for replicate samples was 99.5%. The average percentage of successful genotyping was 98.5%.
Statistical Methods
We evaluated the relationship between IGF1 haplotypes and single variants with prostate cancer risk. Haplotype frequencies among case patients with prostate cancer and control subjects were estimated by using genotype data of the tagging SNPs, as described by Stram et al. (22). Haplotype dosage (i.e., an estimate of the number of copies of haplotype h) for each individual and for each haplotype h was computed by use of that individual's genotype data, and haplotype frequency estimates were obtained from the E-M algorithm (23). A likelihood ratio test was performed to provide a global test for associations with the common haplotypes (i.e., haplotype frequency 
5%) in each block. Odds ratios and 95% confidence intervals for each common haplotype within each block were estimated by unconditional logistic regression with all other observed haplotypes used as the reference category.
To investigate the hypothesis that genetic susceptibility to prostate cancer risk is associated with single causal variants, we evaluated the relationship between IGF1 genotypes and disease risk. By use of the 29 tagging SNPs and genotype data obtained from the multiethnic reference panel, we estimated the alleles of the remaining 35 SNPs that were not genotyped in the case–control study for each individual, which we refer to hereafter as "unmeasured SNPs." Within each region of strong linkage disequilibrium, we used genotype data from an individual tagging SNP or a combination of tagging SNPs to predict each individual's genotype for the unmeasured SNPs by calculating the squared correlations (
) between each SNP (s) and their estimates obtained from the E-M algorithm. Genotypes of SNPs between regions of strong linkage disequilibrium were predicted by including them as part of a neighboring block if they met a minimum of at least .7 for each ethnic group. Odds ratios and 95% confidence intervals were estimated by unconditional logistic regression for the association between IGF1 genotypes and prostate cancer risk. Stepwise logistic regression analysis was conducted to evaluate which combination of SNPs provided the best fit to prostate cancer risk.
We conducted permutation testing to guide interpretation of nominally statistically significant SNP associations. Case–control status within strata of a racial or ethnic group was randomly permuted 5000 times for the 29 tagging SNPs. The smallest observed P value for the associated SNP was examined in relation to the permutation distribution of minimal P values associated with the heterozygote and homozygote mutant variants; the homozygote wild type was used as the reference. For example, if a nominal P value of .05 marked the 25th percentile of this distribution, then the permutation P value would be .25. All P values quoted are from two-sided tests.
|
RESULTS |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
Genetic Characterization of the IGF1 Locus
We first sequenced three of the four IGF1 exons in DNA from 95 patients with advanced prostate cancer. We found no missense polymorphisms in these three exons (exon 2 did not meet our sequencing criteria and was therefore not sequenced). We then genotyped 64 SNPs spanning 156 kb of the IGF1 locus (from 23.4 kb upstream to 47.8 kb downstream) in a multiethnic panel of 349 individuals. We identified the following four regions of strong linkage disequilibrium: block 1 (SNPs 1–9; size = 23 kb) spanned the upstream region of IGF1, block 2 (SNPs 11–17; size = 11 kb) included intron 2, block 3 (SNPs 19–44; size = 60 kb) spanned the majority of the gene from intron 2 through the 3' untranslated region, and block 4 (SNPs 46–62; size = 38 kb) covered the downstream region of IGF1 predominantly (Fig. 1). The distances between adjacent blocks were 7 kb or less.
|
Common haplotypes were inferred within each region of linkage disequilibrium, and their corresponding frequencies are shown in Table 2. In blocks 1 and 2, we observed five and six common haplotypes (haplotype frequency
5%), respectively. For blocks 3 and 4, we observed 12 common haplotypes in each of these blocks. A total of 29 tagging SNPs were able to predict the common haplotypes and 33 of the 35 unmeasured SNPs across the IGF1 locus. The common haplotypes within each ethnic group accounted for 77%–100% of the chromosomes (total of 698 chromosomes; i.e., the summation over all of the common haplotype frequencies within each ethnic group) in the panel population (Table 2).
|
Characteristics of Case–Control Study Subjects
Characteristics of the 2320 case patients and 2290 control subjects are presented in Table 3. The mean age at reference for case patients (i.e., age of diagnosis) or for control subjects (i.e., age at blood draw) was 68.3 or 67.9 years, respectively. Each ethnic group constituted 20%–30% of the study population, with the exception of the Hawaiian group, which constituted only 3%. Case patients were more likely to report a family history of prostate cancer (either affected father or brother) than control subjects (P<.001), as expected. All of the following results were altered only slightly when analyses were adjusted for a family history of prostate cancer, and so the unadjusted values are given below.
|
Case–Control Analysis
We first searched for haplotype associations by global tests of differences in risk. We found statistically significant associations between haplotypes and the risk of prostate cancer in blocks 1 (P = .021) and 2 (P = .044) but not in blocks 3 (P = .374) and 4 (P = .278). At least one haplotype within each block was nominally associated with prostate cancer risk (Table 4). We observed positive associations between prostate cancer risk and haplotype 1B (OR = 1.21, 95% CI = 1.04 to 1.40), haplotype 2C (OR = 1.24, 95% CI = 1.06 to 1.44), haplotype 3C (OR = 1.25, 95% CI = 1.03 to 1.50), or haplotype 4D (OR = 1.19, 95% CI = 1.02 to 1.39). All of these haplotypes were associated with increased risk of prostate cancer across all racial and ethnic groups, with the exception of Hawaiians for whom power was limited by the small sample size (Table 5). The haplotypes associated with risk (i.e., haplotypes 1B, 2C, 3C, and 4D) appeared to represent the same long-range haplotype, indicating that these haplotypes tended to cosegregate with each other and, therefore, were likely to reflect the same underlying signal (Fig. 2).
|
|
|
Thirty-three of the 35 unmeasured SNPs in the case–control study were predicted by the set of 29 tagging SNPs, with an average
of .91. SNP18 (hCV2801091) and SNP63 (rs2971578), which are both in areas of weaker linkage disequilibrium, could not be predicted with an of at least .7 in the five racial or ethnic groups (for SNP18, = .04 to .38; for SNP63, = .17 to .38). The association between prostate cancer risk and the 63 SNPs (29 tagging SNPs, 33 predicted SNPs, and SNP18) is shown in Fig. 3. We were unable to genotype SNP63 because a TaqMan assay could not be manufactured for it. We observed nominally statistically significant associations (Ptrend = .002–.05) between prostate cancer risk and 17 SNPs (Table 6). Among the 17 SNPs, we identified SNP3 (an unmeasured SNP, rs7978742) and SNP4 (a tagging SNP, rs7965399) as being perfectly correlated ( = 1.00, for all five racial or ethnic groups). Stepwise logistic regression analysis showed that these SNPs provided the best fit to prostate cancer risk. No other SNPs gave statistically significant results for prostate cancer risk after adjustment for SNP3 or SNP4. SNP3 and SNP4 were also able to explain our haplotype findings in a similar stepwise logistic regression analysis of SNP3, SNP4, and associated haplotypes.
|
|
SNP3 (rs7978742) and SNP4 (rs7965399), which are located approximately 17 kb upstream of the transcription start site, revealed the strongest nominal associations with prostate cancer risk (Ptrend = .002). The CT genotype for SNP4 was statistically significantly associated with increased risk of prostate cancer, compared with the common homozygous TT genotype (OR = 1.25, 95% CI = 1.09 to 1.43; P = .001). This association was also statistically significant for nonadvanced disease (OR = 1.32, 95% CI = 1.13 to 1.55; P<.001). In addition, a positive association, although not statistically significant, was observed with advanced disease (OR = 1.17, 95% CI = 0.96 to 1.42). Analyses stratified by ethnic group revealed an overall consistent positive pattern across all racial and ethnic groups with the exception of Hawaiians (Table 7). The CC genotype for SNP4 was also positively associated with risk (OR = 1.26, 95% CI = 0.95 to 1.68).
|
To measure how often such results would be expected by chance in a study of these 29 tagging SNPs, we performed a permutation test in which case–control status was randomized within each ethnic group and each SNP was tested for its association with height, which generated a null distribution of P values. Of the 29 tagging SNPs, SNP4 (rs7965399) displayed the smallest observed P value; levels of statistical significance as large as that of SNP4 were observed in 5.6% of the simulated null datasets. This finding indicates that, if a similar study were repeated under a null distribution (i.e., no IGF1 variant associated with prostate cancer), an association similar to that observed with SNP4 would occur by chance 5.6% of the time.
|
DISCUSSION |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
In this study, we identified two perfectly correlated polymorphisms—SNP3 (rs7978742) and SNP4 (rs7965399)—that were statistically significantly associated with an increased risk of prostate cancer. These findings build on previous observations that the IGF pathway appears to play an important role in prostate carcinogenesis. A number of studies have found that men with the highest circulating levels of IGF-I, compared with men with the lowest levels, have an increased risk of prostate cancer (9,24,25) and that the IGF-I level in the circulation is a highly heritable trait (26,27). We undertook the first comprehensive evaluation of the relationship between genetic variants in IGF1 and prostate cancer risk in a large multiethnic case–control study. In this report, we identified a set of 29 tagging SNPs that captured most of the common genetic diversity across the locus in a multiethnic population. Our results suggest that inherited variation in IGF1 may play a role in prostate cancer risk.
We identified several haplotypes and SNPs that were associated with an increased risk of prostate cancer. These genetic associations were detected across the entire IGF1 locus (156 kb) and had similar magnitudes of effect (i.e., odds ratios of approximately 1.2). Given the strong correlation observed between neighboring SNPs, the most parsimonious explanation is the existence of one signal that is detected at several sites because of underlying correlation between variants across the locus. The strongest signal was located in block 1, a noncoding region that begins 438 base pairs upstream of the IGF1 transcription site and spans approximately 23 kb. Given the upstream location of this signal, it is possible that a predisposing variant within this region may influence IGF1 promoter and transcriptional activities; additional in vitro studies are needed to examine the functional consequences of such genetic variation. This region of block 1 also contains the (CA)n repeat polymorphism that has been previously proposed to be associated with prostate cancer risk, although studies of this polymorphism have provided mixed results (10–14). By use of data from a prior study (28), we determined that the less common (CA)n repeat length allele was in linkage disequilibrium with the minor C allele of SNP4; however, whether SNP4 or another linked marker is the causal variant remains to be determined.
Our study of 2320 case patients and 2290 control subjects had substantial power to detect modest genetic effects. Our powers to detect the observed odds ratio of 1.25 for the genetic effect and allele frequency of 20% (total population) were 99% and 95% (
= .05, two-sided) under log-additive and dominant models, respectively (Quanto software, Los Angeles, CA) (29). The similar genetic effects observed for both the heterozygote (OR = 1.25) and homozygote (OR = 1.26) classes of SNP4 suggest that SNP4 may operate under a dominant model, although other models may also be consistent with these data.
For men in our cohort who carried either one or two copies of the SNP4 C allele, the absolute risks of being diagnosed with prostate cancer by the age of 70 years were 24%, 6%, 6%, 11%, and 15% for African Americans, Hawaiians, Japanese, Latinos, and whites, respectively. Further, approximately 10% of the cases of prostate cancer in this study could be attributed to carrying a SNP4 variant. In other words, if the population contained only individuals with the wild-type SNP4 T allele, the prevalence of prostate cancer would be 10% lower. In contrast to Mendelian cancer syndromes (such as BRCA1 and BRCA2), in which inherited mutations confer a very high risk of disease in a small fraction of the population (1%–2%) (30), most of the genetic liability of cancer cases is believed to be associated with variants that modestly elevate risk and are present in a large proportion of the population (31).
A strength of our study was the ability to test variation across large and diverse populations of 500–700 African American, Japanese, Latino, and white case patients with prostate cancer. We observed consistent genetic effects across populations, suggesting that the inherited variation in IGF1 behaves similarly among ancestral groups and shares an overall biologic effect. Although ethnicity is an established risk factor for prostate cancer, with African Americans having the highest rates of disease, followed by whites, Latinos, and Asian/Pacific Islanders (32), inherited differences in IGF1 alleles cannot account for the variation in risk across populations. This result is consistent with studies of circulating levels of IGF-I, in which the variations in IGF-I levels do not mirror the racial or ethnic patterns of disease (33–35).
The similar genetic effects of the SNPs across racial or ethnic groups do not preclude the possibility that environmental factors and other genetic variants may modify the biologic effects of IGF1 genetic variation. For example, nutritional factors are important regulators of circulating IGF-I levels (36), and recent studies have demonstrated that dietary factors, such as protein and dairy intake, are associated with higher levels of circulating IGF-I (37–39). It is possible that the genetic effects of IGF1 may be stronger in the presence of dietary factors that promote higher levels of IGF-I and weaker in the absence of such nutrients. Additional studies are needed to examine whether such interactive effects between IGF1 and environmental and/or other genetic variants may contribute to differences in prostate cancer susceptibility across groups.
We investigated whether the genetic effects of IGF1 displayed heterogeneity by age of onset and found similar results for both early age of diagnosis (<60 years) and late age of diagnosis (>77 years) (data not shown). We also examined the potential bias in case ascertainment if the predisposing variant had greater penetrance among those who were diagnosed as a result of intensive screening because of a positive family history (a situation that would lead to overestimation of the genetic effects). We observed similar frequencies of the CT and CC genotypes for SNP4 among case patients who reported a family history of prostate cancer (CT = 30% and CC = 5%) and those who did not (CT = 29% and CC = 6%), indicating no evidence of this potential bias.
Our study has several limitations. Because we did not type all of the polymorphisms at the IGF1 locus but, instead, relied on the inherent linkage disequilibrium at this location, our study could identify areas of interest but could not pinpoint a causal variant. Extensive resequencing of this locus in case patients and control subjects may help to identify a causal variant. In addition, our study cannot exclude the possibility of an association between rare IGF1 variants and prostate cancer risk.
Any claim that inherited variation in a region is associated with a disease must be carefully scrutinized (40,41). Follow-up studies that attempt to replicate initial positive reports are often costly and underpowered to detect effects of modest magnitude. Failing to observe consistent associations across studies generates uncertainty about the role of a locus in disease pathogenesis (17). For genetic association studies, the traditional threshold of statistical significance, a P value of .05, may not be sufficiently stringent given the low prior probability that any of the millions of variants in the human genome plays a role in disease. Overly liberal thresholds can lead to false-positive results when the number of tests (i.e., SNPs) is large (thousands to millions) in relation to the true underlying signal (more likely on the scale of tens to hundreds). This problem has probably contributed to difficulties in replicating results in subsequent studies (42,43).
Attempts to distinguish between true-positive and false-positive results are usually predicated on a combination of the prior likelihood of the result being true and the magnitude of the observed P value. Individual studies are rarely large enough to detect with a high degree of statistical confidence the often modest risks associated with complex disease traits (44,45). Therefore, in evaluating a nominally statistically significant result, it is essential to address the possible sources of false-positive results. False-positive results can arise from statistical fluctuations, issues surrounding multiple testing, and population stratification. Attempts to address these issues can help guide the interpretation of suggestive findings.
For our study, we evaluated the statistical significance of the associations with 29 tagging SNPs by subjecting our nominally statistically significant associations to permutation testing (46). Permutation testing is generally used to empirically assess the statistical significance of a nominally statistically significant P value and to guide the interpretation of results when multiple hypotheses are tested. The permutation P value for our best SNP was .056. This result indicates that, when considering IGF1 as an experimental unit, a similarly strong result would be obtained by chance only 5.6% of the time.
In addition, we also applied the false-positive report probability proposed by Wacholder et al. (47) to our results. The false-positive report probability estimates the likelihood that a result is false positive by incorporating the prior probability that the locus is involved in the disease and the estimated magnitude of the effect size of the variant (47). By use of this framework, the model then evaluates the likelihood of a false-positive result given the observed odds ratio and confidence interval. Although there is an inherent subjectivity to this process, a range of likelihoods (and therefore a range of the probability of a false-positive result) may help to guide the investigator's priorities regarding further pursuit of the variant(s) under study. From published data on IGF-I levels and prostate cancer, we believe an odds ratio of 1.50 is a reasonable magnitude of effect; for a prior probability of .01, .001, or .0001 the likelihood that the observed effect of SNP4 is false positive is estimated to be 10%, 54%, and 92%, respectively.
Population stratification is unlikely to be operative in our study, given that we observed a consistent trend across all racial and ethnic groups (with the exception of Hawaiians, for whom the sample size was small) for the SNPs that demonstrated a positive association with prostate cancer risk. The likelihood that stratification would occur in the same direction across all groups is low. Furthermore, the strongest effect observed was among whites, a group that may have a lower likelihood of stratification than other groups (48,49).
Prior evidence implicating the product of the IGF1 locus in prostate cancer pathogenesis (9,24,25), coupled with the results from this study, provide a solid foundation for attempting to replicate these findings in other cohorts, such as the National Cancer Institute–sponsored Cohort Consortium (http://epi.grants.cancer.gov/BPC3/). With further investigation of the IGF1 system and the identification of other genetic variants that contribute to prostate cancer risk, the variants so identified may be useful for prostate cancer screening and prevention. Given that IGF-I levels have been found to be highly heritable and related to prostate cancer risk, it should be of particular interest to examine whether genetic variation in IGF1 can account for some of the variation in circulating IGF-I levels (9,24,25,27,50). By identifying the mechanisms in which inherited differences in IGF1 influence disease, we will further advance our understanding of prostate cancer biology and disease susceptibility.
|
NOTES |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
Supported by grants CA 63464 and CA 54281 National Cancer Institute, National Institutes of Health, Department of Health and Human Services. M.L. Freedman is supported by a Howard Hughes Medical Institute physician postdoctoral fellowship and Department of Defense Health Disparity Training-Prostate Scholar Award DAMD 17-02-1-0246.
We are indebted to the participants of the Multiethnic Cohort, who have contributed to a better understanding of the genetic contributions to cancer susceptibility. We thank Noel Burtt, Loreall Pooler, Stephanie Riley, and David Wong for their laboratory assistance and Celeste L Pearce, Chris Haiman, Katherine DeLellis, Kris Monroe, Hank Huang, and Peggy Wan for their technical support.
|
REFERENCES |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
(1) Jones JI, Clemmons DR. Insulin-like growth factors and their binding proteins: biological actions. Endocr Rev 1995;16:3–34.[CrossRef][ISI][Medline]
(2) Pietrzkowski Z, Mulholland G, Gomella L, Jameson BA, Wernicke D, Baserga R. Inhibition of growth of prostatic cancer cell lines by peptide analogues of insulin-like growth factor 1. Cancer Res 1993;53:1102–6.
(3) Cohen P, Peehl DM, Lamson G, Rosenfeld RG. Insulin-like growth factors (IGFs), IGF receptors, and IGF-binding proteins in primary cultures of prostate epithelial cells. J Clin Endocrinol Metab 1991;73:401–7.[Abstract]
(4) Ngo TH, Barnard RJ, Leung PS, Cohen P, Aronson WJ. Insulin-like growth factor I (IGF-I) and IGF binding protein-1 modulate prostate cancer cell growth and apoptosis: possible mediators for the effects of diet and exercise on cancer cell survival. Endocrinology 2003;144:2319–24.
(5) Kaplan PJ, Mohan S, Cohen P, Foster BA, Greenberg NM. The insulin-like growth factor axis and prostate cancer: lessons from the transgenic adenocarcinoma of mouse prostate (TRAMP) model. Cancer Res 1999;59:2203–9.
(6) Ruan W, Powell-Braxton L, Kopchick JJ, Kleinberg DL. Evidence that insulin-like growth factor I and growth hormone are required for prostate gland development. Endocrinology 1999;140:1984–9.
(7) DiGiovanni J, Kiguchi K, Frijhoff A, Wilker E, Bol DK, Beltran L, et al. Deregulated expression of insulin-like growth factor 1 in prostate epithelium leads to neoplasia in transgenic mice. Proc Natl Acad Sci USA 2000;97:3455–60.
(8) Renehan AG, Zwahlen M, Minder C, O'Dwyer ST, Shalet SM, Egger M. Insulin-like growth factor (IGF)-I, IGF binding protein-3, and cancer risk: systematic review and meta-regression analysis. Lancet 2004;363:1346–53.[CrossRef][ISI][Medline]
(9) Chan JM, Stampfer MJ, Ma J, Gann P, Gaziano JM, Pollak M, et al. Insulin-like growth factor-I (IGF-I) and IGF binding protein-3 as predictors of advanced-stage prostate cancer. J Natl Cancer Inst 2002;94:1099–106.
(10) Nam RK, Zhang WW, Trachtenberg J, Jewett MAS, Emami M, Vesprini D, et al. Comprehensive assessment of candidate genes and serological markers for the detection of prostate cancer. Cancer Epidemiol Biomarkers Prev 2003;12:1429–37.
(11) Tsuchiya N, Wang L, Horikawa Y, Inoue T, Kakinuma H, Matsuura S, et al. CA repeat polymorphism in the insulin-like growth factor-I gene is associated with increased risk of prostate cancer and benign prostatic hyperplasia. Int J Oncol 2005;26:225–31.[ISI][Medline]
(12) Schildkraut JM, Demark-Wahnefried W, Wenham RM, Grubber J, Jeffreys AS, Grambow SC, et al. IGF1 (CA)19 repeat and IGFBP3-202 A/C genotypes and the risk of prostate cancer in black and white men. Cancer Epidemiol Biomarkers Prev 2005;14:403–8.
(13) Li L, Cicek MS, Casey G, Witte JS. No association between genetic polymorphisms in IGF-I and IGFBP-3 and prostate cancer. Cancer Epidemiol Biomarkers Prev 2004;13:497–8.
(14) Neuhausen SL, Slattery ML, Garner CP, Ding YC, Hoffman M, Brothman AR. Prostate cancer risk and IRS1, IRS2, IGF1, and INS polymorphisms: strong association of IRS1 G972R variant and cancer risk. Prostate 2005;64:168–74.[CrossRef][ISI][Medline]
(15) Kolonel LN, Henderson BE, Hankin JH, Nomura AM, Wilkens LR, Pike MC, et al. A multiethnic cohort in Hawaii and Los Angeles: baseline characteristics. Am J Epidemiol 2000;151:346–57.
(16) Nickerson DA, Tobe VO, Taylor SL. PolyPhred: automating the detection and genotyping of single nucleotide substitutions using fluorescence-based resequencing. Nucleic Acids Res 1997;25:2745–51.
(17) Freedman ML, Pearce CL, Penney KL, Hirschhorn JN, Kolonel LN, Henderson BE, et al. Systematic evaluation of genetic variation at the androgen receptor locus and risk of prostate cancer in a multiethnic cohort study. Am J Hum Genet 2005;76:82–90.[CrossRef][ISI][Medline]
(18) Haiman CA, Stram DO, Pike MC, Kolonel LN, Burtt NP, Altshuler D, et al. A comprehensive haplotype analysis of CYP19 and breast cancer risk: the Multiethnic Cohort. Hum Mol Genet 2003;12:2679–92.
(19) Lewontin RC. The interaction of selection and linkage. II. Optimum models. Genetics 1964;50:757–82.
(20) Gabriel SB, Schaffner SF, Nguyen H, Moore JM, Roy J, Blumenstiel B, et al. The structure of haplotype blocks in the human genome. Science 2002;296:2225–9.
(21) Excoffier L, Slatkin M. Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol Biol Evol 1995;12:921–7.[Abstract]
(22) Stram DO, Haiman CA, Hirschhorn JN, Altshuler D, Kolonel LN, Henderson BE, 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.[CrossRef][ISI][Medline]
(23) Zaykin DV, Westfall PH, Young SS, Karnoub MA, Wagner MJ, Ehm MG. Testing association of statistically inferred haplotypes with discrete and continuous traits in samples of unrelated individuals. Hum Hered 2002;53:79–91.[ISI][Medline]
(24) Stattin P, Rinaldi S, Biessy C, Stenman UH, Hallmans G, Kaaks R. High levels of circulating insulin-like growth factor-I increase prostate cancer risk: a prospective study in a population-based nonscreened cohort. J Clin Oncol 2004;22:3104–12.
(25) Harman SM, Metter EJ, Blackman MR, Landis PK, Carter HB, Baltimore Longitudinal Study of Aging. Serum levels of insulin-like growth factor I (IGF-I), IGF-II, IGF-binding protein-3, and prostate-specific antigen as predictors of clinical prostate cancer. J Clin Endocrinol Metab 2000;85:4258–65.
(26) Harrela M, Koistinen H, Kaprio J, Lehtovirta M, Tuomilehto J, Eriksson 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.[ISI][Medline]
(27) Hong Y, Pedersen NL, Brismar K, Hall K, de Faire U. Quantitative genetic analyses of insulin-like growth factor I (IGF-I), IGF-binding protein-1, and insulin levels in middle-aged and elderly twins. J Clin Endocrinol Metab 1996;81:1791–7.[Abstract]
(28) DeLellis K, Ingles S, Kolonel L, McKean-Cowdin R, Henderson B, Stanczyk F, et al. IGF1 genotype, mean plasma level and breast cancer risk in the Hawaii/Los Angeles multiethnic cohort. Br J Cancer 2003;88:277–82.[CrossRef][ISI][Medline]
(29) Gauderman WJ. Sample size requirements for association studies of gene-gene interaction. Am J Epidemiol 2002;155:478–84.
(30) Easton DF. How many more breast cancer predisposition genes are there? Breast Cancer Res 1999;1:14–7.[CrossRef][Medline]
(31) Balmain A, Gray J, Ponder B. The genetics and genomics of cancer. Nat Genet 2003;33 Suppl:238–44.
(32) Ries LAG, Eisner MP, Kosary CL, Hankey BF, Miller BA, Clegg L, et al. (eds). SEER Cancer Statistics Review, 1975–2002, National Cancer Institute. Bethesda, MD. Available at: http://seer.cancer.gov/csr/1975_2002 [Last accessed: December 14, 2005.]
(33) DeLellis K, Rinaldi S, Kaaks RJ, Kolonel LN, Henderson B, Le Marchand L. Dietary and lifestyle correlates of plasma insulin-like growth factor-I (IGF-I) and IGF binding protein-3 (IGFBP-3): the multiethnic cohort. Cancer Epidemiol Biomarkers Prev 2004;13:1444–51.
(34) Platz EA, Pollak MN, Rimm EB, Majeed N, Tao Y, Willett WC, et al. Racial variation in insulin-like growth factor-1 and binding protein-3 concentrations in middle-aged men. Cancer Epidemiol Biomarkers Prev 1999;8:1107–10.
(35) Winter DL, Hanlon AL, Raysor SL, Watkins-Bruner D, Pinover WH, Hanks GE, et al. Plasma levels of IGF-1, IGF-2, and IGFBP-3 in white and African-American men at increased risk of prostate cancer. Urology 2001;58:614–8.[CrossRef][ISI][Medline]
(36) Thissen JP, Ketelslegers JM, Underwood LE. Nutritional regulation of the insulin-like growth factors. Endocr Rev 1994;15:80–101.[Abstract]
(37) Giovannucci E, Pollak M, Liu Y, Platz EA, Majeed N, Rimm EB, et al. Nutritional predictors of insulin-like growth factor I and their relationships to cancer in men. Cancer Epidemiol Biomarkers Prev 2003;12:84–9.
(38) Holmes MD, Pollak MN, Willett WC, Hankinson SE. Dietary correlates of plasma insulin-like growth factor I and insulin-like growth factor binding protein 3 concentrations. Cancer Epidemiol Biomarkers Prev 2002;11:852–61.
(39) Heaney RP, McCarron DA, Dawson-Hughes B, Oparil S, Berga SL, Stern JS, et al. Dietary changes favorably affect bone remodeling in older adults. J Am Diet Assoc 1999;99:1228–33.[CrossRef][ISI][Medline]
(40) Page GP, George V, Go RC, Page PZ, Allison DB. "Are we there yet?": Deciding when one has demonstrated specific genetic causation in complex diseases and quantitative traits. Am J Hum Genet 2003;73:711–9.[CrossRef][ISI][Medline]
(41) Lander ES, Schork NJ. Genetic dissection of complex traits. Science 1994;265:2037–48.
(42) Hirschhorn JN, Lohmueller K, Byrne E, Hirschhorn K. A comprehensive review of genetic association studies. Genet Med 2002;4:45–61.[ISI][Medline]
(43) Ioannidis JP, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG. Replication validity of genetic association studies. Nat Genet 2001;29:306–9.[CrossRef][ISI][Medline]
(44) Altshuler D, Hirschhorn JN, Klannemark M, Lindgren CM, Vohl MC, Nemesh J, et al. The common PPAR
Pro12Ala polymorphism is associated with decreased risk of type 2 diabetes. Nat Genet 2000;26:76–80.[CrossRef][ISI][Medline]
(45) The CHEK2 Breast Cancer Case-Control Consortium. CHEK2*1100delC and susceptibility to breast cancer: a collaborative analysis involving 10,860 breast cancer cases and 9,065 controls from 10 studies. Am J Hum Genet 2004;74:1175–82.[CrossRef][ISI][Medline]
(46) Hirschhorn JN, Daly MJ. Genome-wide association studies for common diseases and complex traits. Nat Rev Genet 2005;6:95–108.[ISI][Medline]
(47) 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.
(48) Wacholder S, Rothman N, Caporaso N. Population stratification in epidemiologic studies of common genetic variants and cancer: quantification of bias. J Natl Cancer Inst 2000;92:1151–8.
(49) Freedman ML, Reich D, Penney KL, McDonald GJ, Mignault AA, Patterson N, et al. Assessing the impact of population stratification on genetic association studies. Nat Genet 2004;36:388–93.[CrossRef][ISI][Medline]
(50) Pollak MN, Schernhammer ES, Hankinson SE. Insulin-like growth factors and neoplasia. Nat Rev Cancer 2004;4:505–18.[CrossRef][ISI][Medline]
Manuscript received June 1, 2005; revised November 2, 2005; accepted December 1, 2005.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  V. M. Chia, L. C. Sakoda, B. I. Graubard, M. V. Rubertone, S. J. Chanock, R. L. Erickson, and K. A. McGlynn Risk of Testicular Germ Cell Tumors and Polymorphisms in the Insulin-Like Growth Factor Genes Cancer Epidemiol. Biomarkers Prev., March 1, 2008; 17(3): 721 - 726. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. D. Ryan and P. E. Goss The Emerging Role of the Insulin-Like Growth Factor Pathway as a Therapeutic Target in Cancer Oncologist, January 1, 2008; 13(1): 16 - 24. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. C van der Pols, C. Bain, D. Gunnell, G. Davey Smith, C. Frobisher, and R. M Martin Childhood dairy intake and adult cancer risk: 65-y follow-up of the Boyd Orr cohort Am. J. Clinical Nutrition, December 1, 2007; 86(6): 1722 - 1729. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. Hernandez, C. Grenade, E. R. Santos, C. Bonilla, C. Ahaghotu, and R. A. Kittles IGF-1 and IGFBP-3 gene variants influence on serum levels and prostate cancer risk in African-Americans Carcinogenesis, October 1, 2007; 28(10): 2154 - 2159. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. N Pollak Insulin, insulin-like growth factors, insulin resistance, and neoplasia Am. J. Clinical Nutrition, September 1, 2007; 86(3): 820S - 821S. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. Cheng, K. DeLellis Henderson, C. A. Haiman, L. N. Kolonel, B. E. Henderson, M. L. Freedman, and L. Le Marchand Genetic Determinants of Circulating Insulin-Like Growth Factor (IGF)-I, IGF Binding Protein (BP)-1, and IGFBP-3 Levels in a Multiethnic Population J. Clin. Endocrinol. Metab., September 1, 2007; 92(9): 3660 - 3666. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. A. Savage, K. Woodson, E. Walk, W. Modi, J. Liao, C. Douglass, R. N. Hoover, S. J. Chanock, and The National Osteosarcoma Etiology Study Group Analysis of Genes Critical for Growth Regulation Identifies Insulin-like Growth Factor 2 Receptor Variations with Possible Functional Significance as Risk Factors for Osteosarcoma Cancer Epidemiol. Biomarkers Prev., August 1, 2007; 16(8): 1667 - 1674. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. M. Sinilnikova, J. D. McKay, S. V. Tavtigian, F. Canzian, D. DeSilva, C. Biessy, S. Monnier, L. Dossus, C. Boillot, L. Gioia, et al. Haplotype-Based Analysis of Common Variation in the Acetyl-CoA Carboxylase {alpha} Gene and Breast Cancer Risk: A Case-Control Study Nested within the European Prospective Investigation into Cancer and Nutrition Cancer Epidemiol. Biomarkers Prev., March 1, 2007; 16(3): 409 - 415. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. D. McKay, R. Kaaks, M. Johansson, C. Biessy, F. Wiklund, K. Balter, H.-O. Adami, C. Boillot, L. Gioia-Patricola, F. Canzian, et al. Haplotype-Based Analysis of Common Variation in the Growth Hormone Receptor Gene and Prostate Cancer Risk Cancer Epidemiol. Biomarkers Prev., January 1, 2007; 16(1): 169 - 173. [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] |