Assessing the causal and independent impact of parity-related reproductive factors on risk of breast cancer subtypes

Abstract

Background: Observational evidence proposes a protective effect of having children and an early first pregnancy on breast cancer development; however, the causality of this association remains uncertain. Here, we assess whether parity-related reproductive factors impact breast cancer risk independently of each other and other causally related or genetically correlated factors: adiposity, age at menarche, and age at menopause.

Methods: We used genetic data from UK Biobank for reproductive factors and adiposity, and the Breast Cancer Association Consortium for risk of overall, estrogen receptor (ER) positive and negative breast cancer, and breast cancer subtypes. We applied univariable and multivariable Mendelian randomization (MR) to estimate genetically predicted direct effects of ever parous status, ages at first birth and last birth, and number of births on breast cancer risk.

Results: We found limited evidence for a genetically predicted protective effect of an earlier age at first birth on breast cancer risk. While the univariable analysis revealed later age at first birth decreased ER-negative breast cancer risk (odds ratio (OR): 0.76; 95% confidence interval: 0.61, 0.95 per standard deviation (SD) increase in age at first birth), this effect attenuated with separate adjustment for age at menarche (potential confounder) (OR: 0.83; 0.62, 1.06) and age at menopause (genetically correlated factor) (OR: 0.80; 0.66, 1.01). Furthermore, we found evidence that a later age at first birth decreased HER2-enriched breast cancer risk but only after adjusting for number of births (potential mediator) (OR: 0.28; 0.11, 0.57 per SD increase in age at first birth). In the multivariable analysis, we found little evidence for genetically predicted effects of ever-parous status, age at last birth, or number of births on breast cancer risk; however, analyses of ever-parous status and age at last birth were limited by weak instruments.

Conclusions: This study found minimal evidence of a genetically predicted protective effect of earlier age at first birth on breast cancer risk, while identifying some evidence for a genetically predicted adverse effect on ER-negative breast cancer risk. However, weak instruments limited the multivariable analysis of ever parous status and age at last birth, which may be improved with larger sample sizes.

Keywords: BCAC; Breast cancer risk; Mendelian randomization; Reproductive factors; UK Biobank.

Fig. 1

Causal relationships estimated using Mendelian randomization in previous work (Vabistsevits et al. (2022) [19], Prince et al. (2022) [21], Prince et al. (2023) [22], Richardson et al. (2020) [42]). OR, odds ratio; 95% confidence interval. The width of the line/arrowhead indicates the magnitude of the estimate. Given ever parous status and breast cancer risk are binary traits, odds ratios are presented where this trait is an outcome. Estimates are univariable unless indicated otherwise. * indicates multivariable analyses adjusted for adulthood body size. ** indicates multivariable analysis adjusted for childhood body size

Fig. 2

Variables included in multivariable Mendelian randomization analyses

Fig. 3

Genetic correlation between primary exposures, adjustment variables, and breast cancer risk. * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001

Fig. 4

Univariable and multivariable Mendelian randomization assessing the effects of A ever parous status, B age at first birth, C age at last birth, and D number of births on overall, ER-positive, and ER-negative breast cancer risk. No adjustment indicated findings from the univariable analysis. Effect of ever parous status is presented per log odds of ever parous status, and effects of age at first birth, age at last birth, and number of births are presented per standard deviation increase in the phenotypic exposure. * Initial MVMR results are presented due to weak instruments preventing the use of MVMR estimated using Q-statistic minimization