ABSTRACT
BACKGROUND: Menopausal hormone therapy (MHT) is a risk factor for breast cancer (BC). Evidence suggests that its effect on BC risk could be partly mediated by mammographic density. The aim of this study was to investigate the relationship between MHT, mammographic density and BC risk using data from a prospective study. METHODS: We used data from a case-control study nested within the French cohort E3N including 453 cases and 453 matched controls. Measures of mammographic density, history of MHT use during follow-up and information on potential confounders were available for all women. The association between MHT and mammographic density was evaluated by linear regression models. We applied mediation modelling techniques to estimate, under the hypothesis of a causal model, the proportion of the effect of MHT on BC risk mediated by percent mammographic density (PMD) for BC overall and by hormone receptor status. RESULTS: Among MHT users, 4.2% used exclusively oestrogen alone compared with 68.3% who used exclusively oestrogens plus progestogens. Mammographic density was higher in current users (for a 60-year-old woman, mean PMD 33%; 95% CI 31 to 35%) than in past (29%; 27 to 31%) and never users (24%; 22 to 26%). No statistically significant association was observed between duration of MHT and mammographic density. In past MHT users, mammographic density was negatively associated with time since last use; values similar to those of never users were observed in women who had stopped MHT at least 8 years earlier. The odds ratio of BC for current versus never MHT users, adjusted for age, year of birth, menopausal status at baseline and BMI, was 1.67 (95% CI, 1.04 to 2.68). The proportion of effect mediated by PMD was 34% for any BC and became 48% when the correlation between BMI and PMD was accounted for. These effects were limited to hormone receptor-positive BC. CONCLUSIONS: Our results suggest that, under a causal model, nearly half of the effect of MHT on hormone receptor-positive BC risk is mediated by mammographic density, which appears to be modified by MHT for up to 8 years after MHT termination.
Subject(s)
Breast Density/drug effects , Breast Neoplasms/etiology , Hormone Replacement Therapy/adverse effects , Menopause , Aged , Breast Neoplasms/metabolism , Case-Control Studies , Hormone Replacement Therapy/statistics & numerical data , Humans , Mediation Analysis , Middle Aged , Odds Ratio , Prospective Studies , Receptors, Estrogen/metabolism , Receptors, Progesterone/metabolism , Risk FactorsABSTRACT
OBJECTIVE: Assessing the statistical power to detect susceptibility variants plays a critical role in genome-wide association (GWA) studies both from the prospective and retrospective point of view. Power is empirically estimated by simulating phenotypes under a disease model H1. For this purpose, the gold standard consists in simulating genotypes given the phenotypes (e.g. Hapgen). We introduce here an alternative approach for simulating phenotypes under H1 that does not require generating new genotypes for each simulation. METHODS: In order to simulate phenotypes with a fixed total number of cases and under a given disease model, we suggest 3 algorithms: (1) a simple rejection algorithm; (2) a numerical Markov chain Monte-Carlo (MCMC) approach, and (3) an exact and efficient backward sampling algorithm. In our study, we validated the 3 algorithms both on a simulated dataset and by comparing them with Hapgen on a more realistic dataset. For an application, we then conducted a simulation study on a 1000 Genomes Project dataset consisting of 629 individuals (314 cases) and 8,048 SNPs from chromosome X. We arbitrarily defined an additive disease model with two susceptibility SNPs and an epistatic effect. RESULTS: The 3 algorithms are consistent, but backward sampling is dramatically faster than the other two. Our approach also gives consistent results with Hapgen. Using our application data, we showed that our limited design requires a biological a priori to limit the investigated region. We also proved that epistatic effects can play a significant role even when simple marker statistics (e.g. trend) are used. We finally showed that the overall performance of a GWA study strongly depends on the prevalence of the disease: the larger the prevalence, the better the power. CONCLUSIONS: Our approach is a valid alternative to Hapgen-type methods; it is not only dramatically faster but has 2 main advantages: (1) there is no need for sophisticated genotype models (e.g. haplotype frequencies, or recombination rates), and (2) the choice of the disease model is completely unconstrained (number of SNPs involved, gene-environment interactions, hybrid genetic models, etc.). Our 3 algorithms are available in an R package called 'waffect' ('double-u affect', for weighted affectations).