Biomarkers are important tools in epidemiologic research. They can be used to elucidate the etiology of a disease and in risk prediction modeling. However, the large study population, long follow-up time for sufficient case accrual, expense and logistics of sample collection and biomarker measurement are substantial hurdles to overcome in biomarker studies. Collaborations with multiple large cohort studies allow for pooling data to increase analytic sample size and statistical power. Thus we have established a collaboration between four prospective cohort studies: the Nurses’ Health Study, the Generations Study, the Mayo Mammography Health Study, and the Melbourne Collaborative Cohort Study in order to pool data and resources to establish a large dataset with questionnaire, plasma hormone, mammographic density, and genotype data. In Chapter 1, we conducted an analysis on the association between plasma c-peptide and postmenopausal invasive breast cancer using data from the B2RISK consortium. Our study provides, to our knowledge, the largest prospective assessment of this association including 3557 cases and 4825 non-cases. In Chapter 2, we assessed the performance of three existing breast cancer risk prediction models, the Gail model, the Rosner-Colditz model, and the simplified Rosner-Colditz model. We extended these models using biomarker data collected as part of the B2RISK consortium including: a polygenic risk score, percent mammographic density, and plasma hormones. In conclusion, our findings suggest apositive association between plasma c-peptide and postmenopausal breast cancer. This association did not vary by standard breast cancer risk factors of tumor molecular characteristics, nor was this association fully accounted for by the correlation of c-peptide and other hormonal risk factors. We also confirmed and validated, in independent populations, previous analyses that showed large improvements that mammographic density and polygenic risk scores offer to breast cancer risk models. We further determined that plasma estradiol significantly improves the performance of breast cancer risk prediction models in postmenopausal women not taking exogenous hormones.