Study population
Women in this analysis were drawn from the 97,786 female participants in the CPS-II Nutrition Cohort, which was established by the American Cancer Society in 1992 as a subgroup of the larger 1982 CPS-II baseline mortality cohort [
26]. Most participants were aged 50 to 74 years at enrollment in 1992. At baseline they completed a 10-page self-administered questionnaire that included questions on demographic, reproductive, medical, behavioral, environmental, and dietary factors. Beginning in 1997, follow-up questionnaires were sent to cohort members every 2 years to update exposure information and to ascertain newly diagnosed cancers. Incident cancers reported on the questionnaires were verified through medical records, linkage with state cancer registries, or death certificates.
From June 1998 to June 2001, blood samples were collected from a subgroup of 39,376 cohort members (n = 21,965 women). After obtaining informed consent, we collected a maximum of 43 ml of non-fasting whole blood from each participant and samples were separated into aliquots of serum, plasma, red blood cells, and buffy coat. Samples then were frozen in liquid nitrogen vapor phase at about -130°C for long-term storage.
Among the 21,965 women with blood samples who were cancer-free and postmenopausal in 1992, 502 postmenopausal breast cancer cases diagnosed between 1992 and 2001 were reported through follow-up questionnaires and subsequently verified. For all cases, questionnaire information on risk factors for breast cancer was collected before the diagnosis of cancer (that is, in 1992); however, collection of DNA from buffy coat occurred in 1998 to 2001, in some cases after or only slightly before cancer diagnosis. For each case, we randomly selected one cancer-free (except non-melanoma skin cancer) female Nutrition Cohort control matched on single year of age (±6 months), race/ethnicity (white, African American, Hispanic, Asian, other/unknown), date of blood collection (within 6 months), and selected from individuals who were cancer-free (except for non-melanoma skin cancer) at the beginning of the interval preceding the diagnosis of each case using risk-set sampling [
27]. The women who provided blood specimens were similar to the overall population in the distribution of most demographic and lifestyle characteristics.
Statistical analysis
We used a χ2 test to assess whether XRCC1 genotype distributions were in Hardy-Weinberg equilibrium. Conditional logistic regression was used to examine the association between the XRCC1 polymorphisms and breast cancer in two models, one controlling for matching factors only and the other controlling for matching factors and other possible confounders.
Conditional logistic regression modeling was performed because controls were individually matched at the time of selection to each case patient. Potential confounders included in the multivariate models included education (up to and including high school graduate, some college, college graduate or higher), age at menopause (less than 45, 45 to 54, 55 or more years, unknown) and menarche (less than 12/missing, 12 or more years), parity (nulliparous, 1 to 2, 3 or more, unknown), age at first live birth (less than 25, 25 or more, unknown), personal history of benign breast disease (yes, no), family history of breast cancer in mother or sisters (yes, no), alcohol intake (none, not more than one drink per day, more than one drink per day, unknown), postmenopausal hormone replacement use (never, current estrogen replacement therapy, current combination estrogen–progestin replacement therapy, former estrogen replacement therapy, former estrogen–progestin replacement therapy, unknown), smoking status (never, ever, unknown), body mass index (less than 25, 25 to less than 30, 30 or more, unknown), and physical activity (metabolic equivalents (MET-hours) per week) (less than 7, 7 to less than 17.5, 17.5 to less than 31.5, 31.5 or more, unknown). With the exception of age at menopause (8.5% missing), all covariates had less than 3.5% values unknown.
To test whether factors proposed to be associated with oxidative stress modified the association between
XRCC1 polymorphisms and breast cancer risk, we constructed multiplicative interaction terms with cigarette smoking, fruit and vegetable intake, alcohol intake, physical activity, body mass index, folate intake, and multivitamin use. We also tested for potential interaction between the two polymorphisms of interest. Statistical interaction was assessed in multivariate models using the likelihood ratio test, and
p < 0.05 was considered statistically significant [
28].