Dissertations and Theses

Date of Degree


Document Type


Degree Name

Doctor of Philosophy (Ph.D.)


Epidemiology and Biostatistics


Catherine Mary Schooling

Committee Members

Catherine Mary Schooling

Levi Waldron

Chloe A. Teasdale

Mary Beth Terry

Subject Categories

Public Health


Breast cancer, body mass index, testosterone, insulin, mendelian randomization, selection bias


BACKGROUND: Female breast cancer is the most studied cancer, and the second leading cause of cancer death in women, whose aetiology is still not completely clear. The parallel increase of incident obesity and breast cancer in women worldwide raises the question of the role of adiposity, measured by body mass index (BMI), and of major anabolic hormones, such as insulin and testosterone, on breast cancer. Observational studies are open to confounding making it difficult to clearly demonstrate causal effects. Several mendelian randomization (MR) studies have shown inverse association of BMI with breast cancer. To assess selection bias in MR studies of BMI on breast cancer and to obtain less biased estimates, own and proxy reported breast cancer was used in two-sample MR studies to assess the effect of BMI, insulin, and testosterone on breast cancer for European and East Asian women.

METHODS: Genetic instruments from genome-wide association studies (GWAS) of three major exposures: 1) female specific BMI (171,977 women) from the Genetic Investigation of ANthropometric Traits (GIANT) consortium, female specific testosterone (230,454 women) from the UK Biobank, 2) female specific insulin (50,404 women) from the Meta-Analyses of Glucose and Insulin-related traits Consortium (MAGIC)), and 3) liability to breast cancer (Breast Cancer Association Consortium (BCAC), 122,977/105,974 cases/controls) and UK Biobank participant reported sibling breast cancer (16,586/345,223 cases/controls) were applied to a GWAS of maternal lifespan (n=412,937) to assess the harm of these exposures on breast cancer. Then the genetic instruments for BMI, testosterone, insulin were applied to BCAC breast cancer GWAS and participant reported sibling breast cancer, as well as Biobank Japan women’s breast cancer (5,552 cases and 89,731controls). The inverse variance weighted (IVW) method was used as the main analysis. Weighted median and MR-Egger were performed as sensitivity analyses for these univariable MR. Female BMI of East Asian women (n=72,390) from Biobank Japan was used as a positive control since female specific instruments for testosterone or insulin was only available among European women but was applied to assess the effect of testosterone and insulin on breast cancer in East Asian women. Multivariable MR (MVMR) was performed to explore the mechanisms of effect. Multivariable IVW was uses as the main analysis. Multivariable MR-Egger and MR-Lasso were used as sensitivity analyses.

RESULTS: Each genetically predicted standard deviation (s.d.) higher BMI was associated with -1.50 (95% confidence interval (CI): -2.30 to -0.70) years of lifespan. Liability to BCAC breast cancer was less strongly associated with lifespan (-0.53; 95% CI: -0.74 to -0.31) than UK Biobank participant reported sibling breast cancer (-1.74; 95% CI: -2.46 to -1.03). Genetically predicted BMI was inversely associated with breast cancer in BCAC compared with unaffected individuals, odds ratio (OR) 0.74 (95% CI: 0.64 to 0.84) but was unrelated to participant reported sibling breast cancer, OR 0.99 (95% CI: 0.84 to 1.13).

Each s.d. of genetically predicted higher testosterone was positively associated with participant report of sibling breast cancer in the UK Biobank, odds ratio (OR)=1.09 (95% confidence interval (CI): 1.03, 1.15). In East Asians, the association of testosterone with the positive control outcome of BMI was in the expected direction 0.04 standard deviation higher BMI (95% CI: -0.004, 0.08) per s.d. testosterone. Testosterone was also positively associated with breast cancer in East Asian women (OR=1.21, 95% CI: 1.03, 1.42).

Genetically predicted each s.d. higher level of fasting insulin was not associated with participant sibling breast cancer in the UK Biobank, odds ratio (OR)=1.03 (95% confidence interval (CI): 0.86, 1.21). In East Asians, the association of fasting insulin with the positive control outcome of BMI was in the expected direction 0.10 s.d. higher BMI (95% CI: -0.05, 0.25) was resulted from per s.d. higher fasting insulin. Fasting insulin was positively but not significantly associated with breast cancer in East Asian women (OR=1.51, 95% CI: 0.79, 2.87).

The MVMR estimate for testosterone on UK Biobank participant reported sibling breast cancer allowing for BMI (odds ratio (OR)=1.10; 95% confidence interval (CI): 1.02 to 1.17) did not differ from the univariable MR estimate. However, the MVMR estimate for testosterone allowing for BMI (OR=1.15; 95% CI: 0.97 to 1.37) was attenuated compared to the univariable MR estimate (OR=1.21; 95% CI: 1.03 to 1.42) in East Asian women.

CONCLUSION: MR studies of harmful exposures on outcomes that can cause death before recruitment are open to left truncation. Our finding using an outcome GWAS less open to left truncation suggests higher BMI is unlikely to protect against breast cancer. In the absence of definitive MR studies about the role of BMI in breast cancer, the IARC recommendation that higher BMI is a risk factor for post-menopausal breast cancer provides the best guide to policy. Testosterone in women may increase the risk of breast cancer. Whether testosterone operates by increasing estrogen, which is known to cause breast cancer, should be investigated to clarify the causal pathway. These finding may also have important implications for use of endogenous testosterone for the use of exogenous testosterone in women as a treatment for low sexual desire or for use by transgender men. There is not enough evidence to show that insulin in women plays a role in breast cancer in European or East Asian women. Future studies with stronger and larger female-specific and population-specific genetic instruments for fasting insulin, possibly with repeated measures, are needed to evaluate the effect of insulin on breast cancer risk. BMI does not seem to mediate the effect of testosterone on breast cancer in European women but may partly mediate the effect of testosterone on breast cancer in East Asian women. Larger unbiased breast cancer GWAS are needed to confirm this finding. Larger GWAS of fasting insulin or GWAS of a more comprehensive measure of insulin that provide stronger instruments are needed to investigate whether insulin mediates the effect of testosterone on breast cancer, because the current instruments are too weak.

Included in

Public Health Commons



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.