Sellers, Thomas A.

Abstract Population stratification in case-control studies can result from differences in ancestral make-up between cases and controls due to sampling variation. Ancestry is therefore routinely estimated and controlled for in these studies. Although less common, it is also possible for ancestry to truly be associated to the disease of interest. Controlling for ancestry can then compromise the power to detect variants that underlie both ancestry and disease. Using data from a multi-center case-control study of epithelial ovarian cancer (OC) among women of European-American ethnicity in which case-control samples were drawn from four geographically distinct North-American sites, we developed a three-step method to investigate whether the stratification observed across the four sites is due to chance or to a non-spurious association between European ancestry and OC. First, we collapsed genome-wide data into two principal components (PCs) that represent the major axes of European ancestral variation, which correspond to north-south and to west-east geographical clines. Next, we determined the discriminant function based on these two PCs that best separates cases and controls for each individual study site and we used the largest angle between the four discriminant function vectors as a measure of similarity ranging from 0 degrees, where all four vectors are exactly the same, to 180 degrees where two of the vectors face exactly opposite directions. If stratification is due to sampling, it is unlikely that the four sites will present similar discriminant functions by chance alone. We then gauged significance of similarity by permuting the case control status across individuals within each site and recalculating the maximum angle between the discriminant function vectors, thus obtaining a null distribution for the statistic. We found that all four sites present the greatest separation between cases and controls on the west-east axis of ancestral variation, with a higher proportion of ovarian cancer cases presenting western ancestry. The largest angle between the four discriminant function vectors observed was of 36.5 degrees. Permutation testing showed this level of similarity between the functions to have a p-value < 0.01 and suggests that OC risk is somehow tied to European ancestry. It remains to be seen whether shared culture or other exposures such as parity or oral contraceptive use explain this connection or whether genetic variants underlie it and are potentially being missed due to routine ancestry control in association testing. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1678. doi:1538-7445.AM2012-1678

Abstract DNA copy number variations (CNVs) are a significant and ubiquitous source of human genetic variation. However, the influence of CNVs on cancer susceptibility remains poorly understood. Out of 83 genome-wide association studies (GWAS) on cancer to date, only a few studies found significant associations on germline CNVs and cancer. Here we analyzed data from our ongoing two-stage GWAS of four North American case-control studies to test the hypothesis that CNVs in germline DNA from peripheral blood lymphocytes may serve as risk factors for epithelial ovarian cancer (EOC). To reduce disease heterogeneity, we focused on the subset of cases with serous histology. The analysis was therefore based on 942 serous ovarian cancer patients and 1,682 healthy controls who were genotyped using the Illumina 610K quad array. Subjects with extreme noise and genomic waviness in log R ratio (LR) were excluded prior to segmentation; principal component analysis was performed to adjust for batch effects. CNV segmentation was performed on LR data from 22 autosomes using circular binary segmentation embedded in the Copy Number Analysis Module from Golden Helix SNP Variation Suite version 7. Copy number segment covariates were discretized based on the thresholds that signify a transition between copy number states (deletion/no deletion; duplication/no duplication) after the segmentation. Unconditional logistic regression on a log-additive model was used to evaluate the association between copy number states and serous EOC risk after adjusting for study sites. By comparing single marker copy number states at 388,958 SNPs on 22 autosomes, we observed a total of 134 SNPs significantly associated with risk of serous EOC with a p value below 10−6. Associations with deletion polymorphisms were observed on chromosomes 7, 8, 14, and 18 when controlling for false discovery rate at 1%; no duplication polymorphisms were significant. We observed a large deletion at chromosome 14 that occurred in 8.9% of cases but in only 3.9% of controls, with a p value of 5.59×10−8 (Odds Ratio (OR): 2.56, 95% confidence interval (CI): 1.82-3.60). Another common deletion on chromosome 7 occurred in 8.8% of cases and only 4.3% of controls (p=3.83×10−6; OR: 2.23, 95% CI: 1.60-3.10). Two additional regions on chromosomes 8 and 18 with deletion events less than 5% were also identified. Women who harbored the deletion on chromosome 8 were at lower risk of developing serous ovarian cancer (p=4.83×10−8; OR: 0.04, 95% CI: 0.01-0.30). Women who carried the deletion on chromosome 18 had an increased risk for serous EOC (p=9.04×10−7; OR: 7.26, 95% CI: 2.94-17.97). Further validation of these four regions using independent data sets is currently underway. In summary, this is the largest reported genome-wide study of CNVs and serous EOC risk. These preliminary results suggested that germline CNVs may play an important role in ovarian cancer susceptibility. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 1649. doi:1538-7445.AM2012-1649

Abstract Mi(cro)RNAs are short non-coding RNA molecules that play a key role in carcinogenesis by regulating tumor suppressors and oncogenes. Published data has implicated miRNAs in ovarian cancer (OC) development and progression, and we hypothesize that single nucleotide polymorphisms (SNPs) in sites of miRNA: messenger RNA (mRNA) binding may influence OC risk by altering expression levels of targeted mRNAs. To evaluate associations between SNPs in miRNA binding sites and OC risk, we used bioinformatics tools and public databases to identify miRNA binding SNPs relevant to ovarian cancer. We then evaluated the frequency of approximately 1,100 SNPs in ∼20,000 epithelial OC cases and 20,000 controls represented in the international Collaborative Oncological Gene-environment Study. Unconditional logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (CI) between genotypes and case status, with adjustment for the first five principal components representing European ancestry. Log-additive genetic models were applied to each SNP, modeling the number of copies of the minor allele. Subgroup analysis was conducted for serous adenocarcinomas, the most predominant histologic subtype of epithelial OC. Preliminary analysis of 11,215 invasive epithelial OCs and 15,910 controls included four SNPs with P<10−6 and 1 SNP with P<10−5. Analysis of 6,215 invasive serous adenocarcinomas revealed two SNPs with P<10−7 and three SNPs with P<10−6. Noteworthy is rs6104808, a C>T SNP predicted to reside in a highly conserved miRNA binding site of BTBD3 (BTB POZ domain containing 3), a transcription factor suggested to be involved in cell proliferation, cancer progression, and response to platinum-based chemotherapy. rs6104808 was associated with a decreased risk of OC overall (OR (95% CI): 0.42 (0.28-0.62), P=1.26 x 10−5)), and the association was stronger among cases with serous adenocarcinoma (OR (95% CI): 0.30 (0.17-0.50), P=9.89 x 10−6)). This represents the largest, most comprehensive epidemiologic study to date to evaluate associations between miRNA binding site SNPs and OC susceptibility. Although based on a preliminary analysis, these data suggest a significant role and putative biological mechanism for common germline variants in miRNA binding sites leading to cancer development. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 2927. doi:1538-7445.AM2012-2927

Abstract Ovarian cancer is the leading cause of death in gynecologic malignancies in the United States with most patients presenting with late stage disease; patients receive the same treatment regardless of histology or tumor characteristics. The mismatch repair (MMR) molecular pathway is well-defined and known to be involved in up to 20% of ovarian cancers; however, association of deficient v proficient MMR status (dMMR v pMMR status) with risk factors and outcome is not well understood. We are conducting extensive molecular testing of the MMR genes MLH1, MSH2, MSH3, MSH6, and PMS2 on Mayo Clinic invasive epithelial ovarian cancer cases. First, the prevalence of dMMR tumors will be estimated in ovarian cancer cases (n=496) based on loss of expression of MLH1, MSH2, MSH3, MSH6, and PMS2 determined by immunohistochemistry staining of primary antibodies using negative and positive controls. Staining of tissue microarrays (TMAs) is complete; triplicate cores will be scored independently for presence or loss of protein expression by a board-certified pathologist. Second, for those dMMR cases, the possible cause of MMR deficiency will be evaluated considering germline and somatic mutations. The loss of expression of each protein will be characterized as germline if a deleterious mutation is present in peripheral blood or somatic if the pathogenic mutation is only present in tumor tissue. In cases with MLH1 loss somatic hypermethylation will also be considered. Analysis of germline mutation data will include all 496 cases and somatic mutations data will include a subset of cases tested with frozen tumor DNA (n=292) using Fluidigm access array targeted resequencing. Third, the clinical features of dMMR disease will be compared to those of pMMR; in particular, age at diagnosis, histopathology, and disease outcome. Among cases in this analysis, the majority have high-grade serous disease with a mean age at diagnosis of 61 years, about half had recurrence or progression (n=243) of which two-thirds (n=167) are deceased. Finally, known and suspected ovarian cancer risk factors (family history, smoking, body mass index, parity, use of hormone therapy, and genetic variation in susceptibility loci at 2q31, 3q25, 8q24, 9p22, 17q21, and 19p13) will be analyzed by MMR phenotype. Risk factor questionnaire data is available on 462 cases (93%), as well as 737 population controls with germline genome-wide genotype data. About half of the cases (n=231) report a 1st or 2nd degree family member with breast or ovarian cancer. Given the limited knowledge about ovarian cancer etiology and the poor prognosis of advanced disease, it is very important to study risk factors with this molecular subtype. The ovarian cancer dMMR phenotype may have distinct clinicopathological features that could guide risk stratification and be useful in diagnostics and therapeutics. This is the largest study characterizing MMR in an ovarian cancer population to date. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 2619. doi:1538-7445.AM2012-2619

Abstract As pathways of the immune system play an important role in multiple malignancies, we assessed the association between single-nucleotide polymorphisms (SNPs) in inflammation-related genes and risk of ovarian cancer in a multi-site case-control study. Using a custom array, we genotyped 930 epithelial ovarian cancer cases and 1,037 controls for SNPs in ALOX12, ALOX15, ALOX5, CCL11, CCL2, CCL3, CCR3, CRP, CXCL16, IL10, IL15RA, IL18, IL1A, IL1B, IL1RN, IL4R, IL6, IL6R, IL7R, IL8RA, IL8RB, IL9, NOS3, PTGS1, PTGS2, TLR2, and TNF. After analysis by logistic regression, SNPs with p<0.10 were evaluated in an independent data set of 3,143 cases and 2,102 controls from the Follow-up of Ovarian Cancer Genetic Association and Interaction Studies (FOCI) collaboration. Combined analysis revealed association with interleukin 1, alpha (IL1A) SNPs (rs17561 and rs4848300, r2=0.990, p=0.005) which varied by histological subtype (heterogeneity p=0.03). For example, IL1A rs17561, which results in a missense change in amino acid at position 114 and correlates with multiple inflammatory phenotypes, was inversely associated with risk of clear cell, mucinous, and endometrioid subtypes, but not with the serous subtype (Table 1). Additional data on IL1A rs17561 in an even larger collection of ovarian cancer cases and controls are pending and will be analyzed soon. We also found that genotype at rs1864414 in ALOX5 (arachidonate 5-lipoxygenase) was associated with decreased risk combined across subtypes (OR 0.86, 95% CI 0.78-0.95; p=0.002). Thus, inherited variation in IL1A and ALOX5 appears to have a role in ovarian cancer risk, though it is limited to rarer subtypes for IL1A. These data add to existing literature on the importance of inflammation in tumorigenesis and growing evidence of subtype-specific traits in ovarian cancer and may provide important clues on prevention. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 2603. doi:1538-7445.AM2012-2603

Abstract Epithelial ovarian cancer (EOC) is the 6th most commonly diagnosed cancer among women in the world, accounting for 4% of all female cancers. On a worldwide basis, an estimated 204,000 new cases are diagnosed and 125,000 women die of EOC annually, with 5-year survival rate at 45%. Metastasis and its complications are the usual cause of EOC-related deaths. Thus, factors related to tumor aggressiveness, response to therapy, and the underlying health state of the patient often predict survival. Candidate gene and genome-wide association studies (GWAS) for survival from EOC have discovered only a few the genetic variants, accounting for only a small proportion of the heritability. One explanation for the missing heritability is that the common analysis approach _ assessing the effect of each single nucleotide polymorphism (SNP) individually _ is not well suited to the detection of small effects of multiple SNPs. Gene set analysis (GSA) overcomes this limitation by assessing the overall evidence of association of a phenotype with all measured variation in a set of genes or pathway. Therefore, we conducted a GSA using data from an EOC GWAS, using the Illumina 610-quad SNP array, to determine novel gene sets (GSs) associated with overall survival (OS) from EOC (N cases = 1057, N deaths = 495). All subjects were non-Hispanic, non-Jewish Caucasians, with sample and SNP call rates >95%. Subjects with ambiguous gender, unresolved identical genotypes, and less than 80% European ancestry were also excluded. SNPs were mapped to genes within 20 kb, with genes then mapped to GSs defined by the following: KEGG, Gene Ontology (biological, cellular, molecular) and PharmGKB. GSA was completed in which the association of each gene with OS was assessed using a principal component gene-level test (adjusting for study site, age and population sub-structure based on the first eigenvector from EigenStrat), followed by the aggregation of the gene-level p-values using the Gamma method with a soft truncation threshold of 0.15. Empirical p-values were determined using permutation methods. Of the 2,573 GSs assessed, 30 GSs had a p-value < 0.01. The top GSs were related to: mismatch repair (p = 3.0E-4), platelet formation (p=0.001), leukocyte adhesive activation (p=0.002), antigen processing and presentation of endogenous antigens (p = 0.002) or peptide antigens (p = 0.002), platinum pathway (p = 0.003), aging and lifespan (p = 0.004, p = 0.007), immune system development (p = 0.004) and somatic diversification (p=0.007). These results illustrate the usefulness of GSA for the development of novel hypotheses that can be followed-up with additional clinical or functional studies. Moreover, application of GSA to survival from EOC allowed the detection of significant GSs (and the corresponding genes/SNPs within these GSs) that impact OS that were not detected by assessing each SNP individually, where the smallest p-value from single SNP analysis of the GWAS for OS was 1.3E-05. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 2738. doi:10.1158/1538-7445.AM2011-2738

Abstract The etiology of ovarian cancer is poorly understood but there is clearly a heritable component. Efforts to identify susceptibility alleles rarely consider interactions among alleles or their joint effects because it quickly becomes computationally intractable. In this study, we sought to identify multi-SNP effects jointly with a pathway-based analysis (GSEA-SNP) of 1,952 EOC cases and 2,042 frequency-matched controls genotyped with the Illumina 610K array. All subjects were self-reported non-Hispanic non-Jewish Caucasians, with SNP and sample call rates > 95%. Subjects with ambiguous gender, unresolved identical genotypes and < 80% European ancestry were excluded. SNPs with MAF < 1% were excluded. Missing genotypes were inferred using Mach based on the HapMap CEU population. SNPs not within introns were annotated to genes within 100 bp. We retrieved the following databases of gene sets (GSs) from a compiled database, MsigDB: human chromosome and cytogenetic band, chemical and genetic perturbations, canonical pathways, microRNA binding targets, transcription factor targets (TFT), cancer gene neighborhood (CGN), Cancer modules and Gene Ontology. The GSEA-SNP approach calculates and rank orders the trend statistic for association between each SNP and EOC risk. The Enrichment Score (ES) estimates the overrepresentation of top-ranked SNPs for each GS; the statistical significance was estimated using 10,000 permutations. The ES was normalized according to the size of GS to yield Normalized Enrichment Score (NES). The false discovery (FDR) rate was estimated using NES to adjust for multiple hypothesis testing. A total of 5181 gene-sets were included in the analysis. When controlling the FDR at 15%, 14 of the GSs were highly enriched with association signals, including two chromosomal regions, 8q13 (p = 0.0055, FDR = 0.15) and 6p24 (p = 0.0012, FDR = 0.08). BIOCARTA_LYM_PATHWAY, a pathway related to cell adhesion and diapedesis of lymphocytes, was also significant (p = 0.0002, FDR = 0.12). A TFT GS, composed of genes with promoter regions around transcription start sites containing the motif GGCNRNWCTTYS, was associated with risk (p = 0.0004, FDR = 0.08). Currently, no known transcription factors bind to this computationally predicted motif. Molecular function of double stranded RNA binding was also enriched (p = 0.005, FDR = 0.10). The remaining enriched sets were CGN sets, for which the neighborhoods around the cancer genes were originally defined using correlation of gene expression from 4 large data sets of various cancer types. The enriched CGN included: CD48 (p = 0.011), CD53 (p = 0.009), CD97 (p = 0.011), INPP5D (p = 0.010), PTPN6 (p = 0.008), VAV1 (p = 0.011), ITGAL (p = 0.012), PTPRC (p = 0.012), and STAT6 (p = 0.010). In summary, these analyses detected biologically plausible GSs related to etiology of EOC, highlighting SNPs in core enrichment groups that were not identified using individual SNP tests. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 4729. doi:10.1158/1538-7445.AM2011-4729

Abstract Background. The hepatocyte growth factor (HGF) signaling pathway regulates key cellular functions related to tumor invasion and metastasis including cell motility, cell proliferation, and morphogenesis. We previously found that minor alleles at a single-nucleotide polymorphism (SNP; rs5745709) in the gene encoding the cytokine HGF was associated with poorer overall survival in women with invasive epithelial ovarian cancer (EOC) enrolled at Mayo Clinic (MAY) and The Cancer Genome Atlas (TCGA) (HR 1.6, 95% CI 1.3-1.9, p=8.9×10−5). In addition, we found that that genotype correlated with HGF expression in ovarian cancer tissue from TCGA women, suggesting that genetically-regulated expression of HGF may play a role in post-diagnosis survival. To further examine this possibility, we evaluated intensity of antibody immunostaining for HGF, MET (HGF's transmembrane tyrosine kinase receptor), and phospho-MET (pMET; MET's activated form) in tumor cells using tissue microarrays (TMAs) created from genotyped primary peritoneal and EOC cases, and correlated these expression levels with overall survival and HGF SNP genotypes. Methods. Participants were 326 white non-Hispanic women over 20 years of age with histologically-confirmed invasive EOC diagnosed between 1998 and 2009. All cases were enrolled at MAY within one year of diagnosis. Five μm sections of TMAs including three 0.6 mm cores per case were immunostained with HGF, MET, and pMET primary antibodies. Staining levels were classified according to combinations of staining extent and intensity as negative (< 10% cells stained), weak, moderate, or strong. The maximal expression value recorded over the multiple cores for each case was used. Cox proportional hazards regression was used to examine associations of overall survival with HGF, MET and pMET expression levels, adjusted for relevant clinical variables. Associations between expression levels and HGF SNP rs5745709 genotypes were examined using linear regression models. Results. HGF and MET expression were not associated with overall survival in our series of EOC cases (p=0.48 and 0.90, respectively). pMET expression was associated with survival, but in the opposite direction of that hypothesized (HR 0.77, 95% CI 0.39-1.52 for moderate expression; HR 0.55, 95% CI 0.27-1.10 for strong expression; each referent to negative/weak, p=0.01). Expression levels were not significantly associated with HGF rs5745709 genotypes (p=0.80 for HGF expression, p=0.89 for MET, p=0.41 for pMET). Conclusion. HGF expression levels were not associated with overall survival or with HGF SNP genotypes that correlated with survival in a previous study. Functional studies such as this one are crucial as a follow-up to discovery-based studies in order to assess biologic plausibility of observed results. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 893. doi:10.1158/1538-7445.AM2011-893

Abstract Ovarian carcinoma is a significant cause of cancer mortality in women in the United States. Increasing evidence has demonstrated that differential methylation status in lymphocyte DNA may be linked to risk of carcinogenesis. We are investigating the differential DNA methylation in germline DNA obtained from pre-treatment ovarian cancer cases (n=256) and individually-age matched controls (n=274) using the Illumina Infinium Human Methylation 27K BeadChip. We are first studying methylation and the common moderate-risk alleles for ovarian cancer on chromosomes 2q31, 3q25, 8q24, 9p22, 17q21, and 19p13. Within each locus, we correlated single nucleotide polymorphism (SNP) genotypes with methylation values at relevant nearby CpG sites. Preliminary results are available on the first set of 84 cases and 90 controls. At 2q31, cases with TG or TT genotype at rs2072590 were hypomethylated at HOXD1 (a homeobox transcription factor) compared to cases with GG genotype (mean beta TG/TT=0.08, GG=0.09, p=0.034); and at 17q21, cases with AG or GG genotype at rs9303542 were hypomethylated at SKAP1 (a T-cell adaptor protein) compared to cases with AA genotype (mean beta AG/GG=0.14, AA=0.16, p=0.045). At these two loci, similar trends were seen among controls. Logistic regression adjusted for age and methylation chip was then used to assess the association between SNP genotype and ovarian cancer risk, with and without inclusion of the relevant methylation value. At HOXD1 rs2072590, the per-allele ovarian cancer risk of 1.34 attenuated to 1.25 with inclusion of the HOXD1 methylation value as a covariate, suggesting that the disease association may be mediated in part through methylation. Though based on preliminary data, results are consistent with a role for inherited variation on germline methylation at HOXD1 and SKAP1. Analysis of methylation on additional genotyped cases and controls is underway. Finally, because, trans effects may also exist, we are assessing relationships between genotype and methylation at other loci throughout the genome. Such correlative molecular studies have the potential to help elucidate mechanisms underlying associations from genome-wide association studies. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 5616. doi:10.1158/1538-7445.AM2011-5616

Abstract INTRODUCTION: Mammographic breast density is a strong risk factor for breast cancer (BC). We established the Mayo Mammography Health Study (MMHS) cohort study at the Mayo Clinic in Rochester, Minnesota (MN) to examine the association of breast density with BC. We evaluated the influence of the image acquisition technique on the density and BC association. METHODS: From October 2003 to September 2006, all women scheduled for screening mammography at the Mayo Clinic were invited to participate. Eligible women were residents of MN, Iowa (IA) or Wisconsin (WI); age 35+; and had no personal history of BC. A risk factor questionnaire, consent form and permission to link to tumor registries were obtained. Incident BC was identified through 2009 by linkage to the Mayo and state cancer registries. A case-cohort of all incident BCs and 2300 randomly selected women (the subcohort) were used to examine the association of breast density and BC using digitized film mammograms at the time of enrollment while controlling for the influence of the acquisition parameters (peak kilovoltage, milliampere-second, and compressed breast thickness). Two density measures were considered: a quantitative percent density (PD) measure estimated using the computer-assisted thresholding program, Cumulus (University of Toronto), and the 4-category clinical BI-RADS measure. Proportional hazards regression was used to calculate hazards ratios (HR) and 95% confidence intervals (CI) for BC associated with quartiles of PD (0-6.2%,6.3-14.9%, 15.0-25.7% and 25.8%+) and BI-RADS categories 1-4 (almost entirely fatty to extremely dense). All models included age, postmenopausal hormone use (PMH), BMI, and menopausal status. The influence of acquisition parameters was evaluated by examining models with and without their inclusion. RESULTS: A total of 20,982 (50%) women participated in MMHS; 1058 (5%) with a prior history of BC were excluded, for a total cohort of 19,924. Compared to nonresponders, responders were younger (57.5 vs. 58.4 yrs), more likely to have ever used PMH (45% vs. 33%), and more likely to have a BC family history (19% vs 16%). The case-cohort consisted of 317 incident cases and 2300 women in the subcohort. Women were excluded from the current analysis if PD could not be estimated or if acquisition parameters were not available, leaving 249 cases and 1937 in the subcohort. As expected, PD was associated with BC [HR (95% CI): 1.0 (REF), 2.1 (1.4-3.1), 3.0 (2.0-4.5), and 4.6 (3.0-7.0) for quartiles; p-trend<0.001]. Controlling for acquisition parameters attenuated the association [HR (95% CI): 1.0 (REF), 2.3 (1.5-3.4), 2.4 (1.6-3.7), and 3.0 (1.8-5.0) for quartiles; p-trend<0.001]. Results for BI-RADS density were similar to those for PD. CONCLUSION: This study confirms that breast density is a significant risk factor for BC and demonstrates that the acquisition technique confounds the density and BC risk association. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 3716. doi:10.1158/1538-7445.AM2011-3716