We used the polytomous model and the joint Cox model to formally compare the association of BMI with breast cancer by ER and PR status using data from the NIH-AARP Diet and Health Study. The NIH-AARP study was established in 1995 when a baseline screening questionnaire that queried information on medical history and lifestyle characteristics was returned by 566,398 AARP (formerly known as the American Association of Retired Persons) members. Participants were aged 50–71 and resided in 1 of 6 US States (California, Florida, Louisiana, New Jersey, North Carolina and Pennsylvania) or two metropolitan areas (Atlanta, Georgia, and Detroit, Michigan), selected because they were known to have high-quality cancer registries and a large AARP membership (18 (link)).
A total of 54,629 women remained in the analytic cohort after exclusion of pre-menopausal women, women with previous history of cancer as well as current users of hormone therapy and women residing in Florida or Pennsylvania because these two states have substantial proportion of participants with missing ER/PR data. In this cohort, 1,492 cases were diagnosed with breast cancer with known ER/PR status obtained from cancer registry records, in which 246 were classified as ER−/PR− breast cancer cases (16.5%), 231 ER+/PR− cases (15.5%), 18 ER−/PR+ cases (0.01%) and 997 ER+/PR+ cases (66.8%). Body mass index (BMI; kg/m2) was derived from self-reported height and weight on the baseline survey. In statistical models, BMI was expressed as categorical variables (<25.0 kg/m2; 25.0–29.9 kg/m2; ≥30.0 kg/m2). Covariates included in the multivariable models were consistent with those reported in the prior analysis by Ahn et al (18 (link)), namely age, race or ethnic, family history of breast cancer, level of education, age at menarche, age at menopause, age at first birth, parity, smoking status, physical activity, fat intake, alcohol consumption, oophorectomy, and height. The HRs for the association of BMI with the four breast cancer subtypes were estimated simultaneously and the difference in HRs were assessed. The analysis was conducted using SAS 9.2 (SAS Institute, Cary NC).
A total of 54,629 women remained in the analytic cohort after exclusion of pre-menopausal women, women with previous history of cancer as well as current users of hormone therapy and women residing in Florida or Pennsylvania because these two states have substantial proportion of participants with missing ER/PR data. In this cohort, 1,492 cases were diagnosed with breast cancer with known ER/PR status obtained from cancer registry records, in which 246 were classified as ER−/PR− breast cancer cases (16.5%), 231 ER+/PR− cases (15.5%), 18 ER−/PR+ cases (0.01%) and 997 ER+/PR+ cases (66.8%). Body mass index (BMI; kg/m2) was derived from self-reported height and weight on the baseline survey. In statistical models, BMI was expressed as categorical variables (<25.0 kg/m2; 25.0–29.9 kg/m2; ≥30.0 kg/m2). Covariates included in the multivariable models were consistent with those reported in the prior analysis by Ahn et al (18 (link)), namely age, race or ethnic, family history of breast cancer, level of education, age at menarche, age at menopause, age at first birth, parity, smoking status, physical activity, fat intake, alcohol consumption, oophorectomy, and height. The HRs for the association of BMI with the four breast cancer subtypes were estimated simultaneously and the difference in HRs were assessed. The analysis was conducted using SAS 9.2 (SAS Institute, Cary NC).
