The Women’s Contraceptive and Reproductive Experiences Study is a population-based case-control study that recruited women aged 35–64 years old diagnosed with invasive breast cancer between 1994–1998 from five metropolitan areas: Atlanta, Detroit, Los Angeles, Philadelphia, and Seattle. Details of the methods used in this study have been previously published.(13
) Cases were ascertained through Surveillance, Epidemiology, and End Results (SEER) cancer registries at four sites (Atlanta, Detroit, Los Angeles, and Seattle) and by field center staff at one site (Philadelphia). Of 5,982 eligible cases identified, 4,575 (77%) were enrolled and interviewed. Female controls without breast cancer were identified by random digit dialing in the same areas and were frequency matched to cases on 5-year age group, race, and study site. Of the 5,956 eligible controls identified, 4,682 (79%) were enrolled and interviewed. The seven cases and four controls with an unknown history of migraine were excluded from all analyses leaving a total of 4,568 cases and 4,678 controls.
The study protocol was approved by the institutional review board of each participating center, and written informed consent was obtained from all participants. This study excluded all women with a prior history of in situ or invasive breast cancer. All cases and controls were interviewed in-person by trained interviewers using the same questionnaire with standardized probes. Women were asked about exposures occurring before their reference date, which for cases was defined as the date of their breast cancer diagnosis. Reference dates for controls were defined as the dates study staff first contacted a member of prospective controls’ households through random digit dialing. With respect to migraine history, women were asked if “a doctor or other health professional ever told you that you had migraine headaches”, age at migraine diagnosis, and ever use of prescription medications for migraine. In addition, detailed information on other known or suspected breast cancer risk factors, including reproductive history, anthropometric characteristics, use of exogenous hormones, family history of breast cancer, and lifestyle characteristics (including alcohol use and smoking), was collected.
Data on estrogen receptor (ER) status, progesterone receptor (PR) status, and histology of cases were obtained from SEER registry files (Atlanta, Detroit, Los Angeles, and Seattle) or from pathology reports, medical records, and hospital registry abstracts (Philadelphia). We excluded cases with ER−/PR+ tumors (n=185, due to insufficient statistical power) and cases with an unknown ER/PR status (n=804) from our ER/PR analyses. Histology was classified as ductal (n=3,457, ICD-O code 8500), lobular (n=535, ICD-O codes 8520 and 8522), and other (n=576, all other ICD-O codes).
Unconditional logistic regression was used to calculate odds ratios (ORs) and 95% confidence intervals (CI) comparing cases to controls.(14
) We compared cases of each ER/PR status and histological type to controls using polytomous logistic regression.(15
) All analyses were adjusted for age, race, and study site. The referent category was women with no history of clinically diagnosed migraine. Variables considered as potential confounders or effect modifiers included income, education, age at menarche, parity, age at first birth, type of menopause, age at menopause, duration of oral contraceptive use, use of hormone therapy, first-degree family history of breast cancer, body mass index (BMI), smoking status, and average alcohol intake. None of these variables changed our risk estimates by more than 10%, and so none were included as confounders in the final statistical models. Effect modification was assessed using likelihood ratio testing, and none of these variables were observed to be statistically significant effect modifiers (all p-values for interaction>0.05). We also assessed heterogeneity of risk estimates associated with migraine history across breast cancer subtypes defined by ER/PR status and by histology by testing the null hypothesis that the ORs were equivalent to each other (test of homogeneity of ORs). All analyses were conducted using Stata 9.2 (Stata Corp, College Station, TX).