Current risk prediction models estimate the probability of developing breast cancer over a defined period based on information such as family history, non-genetic breast cancer risk factors, genetic information from high and moderate risk breast cancer susceptibility genes and, over the past several years, polygenic risk scores (PRS) from more than 300 common variants. The inclusion of additional data such as PRS improves risk stratification, but it is anticipated that the inclusion of epigenetic marks could further improve model performance accuracy. Here, we present the case for including information on DNA methylation marks to improve the accuracy of these risk prediction models, and consider how this approach contrasts genetic information, as identifying DNA methylation marks associated with breast cancer risk differs inherently according to the source of DNA, approaches to the measurement of DNA methylation, and the timing of measurement. We highlight several DNA-methylation-specific challenges that should be considered when incorporating information on DNA methylation marks into risk prediction models, using BRCA1, a highly penetrant breast cancer susceptibility gene, as an example. Only after careful consideration of study design and DNA methylation measurement will prospective performance of the incorporation of information regarding DNA methylation marks into risk prediction models be valid.