673Is weight more informative than body mass index for breast cancer risk

Background Our previous work on body mass index (BMI) and breast cancer risk found that the association depended on menopausal status but not on familial risk (Hopper, JL., et al, 2018). We now consider whether weight is a more informative risk factor for breast cancer than BMI. Methods We used data from the Prospective Family Study Cohort, a consortium of international prospective cohorts that are enriched for familial risk of breast cancer and include 16,035 unaffected women from 6701 families. Participants were followed for up to 20 years (mean 10.5 years) and there were 896 incident breast cancers with a mean age at diagnosis of 55.7 years. Cox regression was used to model risk associations as a function of age, menopausal status and underlying familial risk. We calculated robust confidence intervals by clustering by family. Model comparisons were made using the Bayesian Information Criterion (BIC). Results In repeating the best-fitting model from our original analyses, but using weight instead of BMI, we found that the log likelihood for the model using weight was 1.92 units greater than for the model using BMI (difference in BIC = 3.84). Therefore, the data are almost 50 times more likely under the model using weight. Conclusions The study found positive evidence that weight gives more information on risk than does BMI. Key messages Analysing breast cancer risk in terms of weight, rather than only BMI, might give greater insight and results that are easier to convey to the public.