We conducted a large population-based case-control study of breast cancer among women 20 to 44 years of age living in the three county Seattle-Puget Sound metropolitan area (King, Pierce, and Snohomish counties) specifically designed to assess the relationship between DMPA use and breast cancer risk.
Cases were women 20 to 44 years old diagnosed with a primary invasive breast cancer between June 2004 and June 2010 with no prior history of in situ or invasive breast cancer. Potentially eligible cases residing in King, Pierce, and Snohomish counties were identified through the Cancer Surveillance System, the population-based tumor registry that serves the 13 counties of western Washington state and participates in the Surveillance, Epidemiology, and End Results program of the National Cancer Institute. Since controls were ascertained via random digit dialing of landline home telephone numbers, to be eligible all cases were also required to have a landline home telephone (160 potentially eligible cases without a landline telephone were identified and excluded). Of the 1,359 eligible cases identified, 1,056 (78%) were interviewed. Of those not enrolled (n=303), 82% refused to be interviewed, 10% could not be located, and 8% died before an interview could be conducted. In addition to basic information on breast cancer diagnosis, we obtained information on tumor characteristics from the cancer registry and from a centralized review of pathology reports. This includes data on estrogen receptor (ER), progesterone receptor (PR), and HER2-neu (HER2) status; and tumor stage, size, and histology.
Using a combination of list-assisted (purchased randomly generated telephone numbers) and Mitosky-Waksberg (telephone numbers randomly generated ourselves using a clustering factor of 5)(8
) random digit dialing methodologies, controls from the general population of female residents of King, Pierce, and Snohomish counties were identified. Controls were frequency matched within 5-year age groups to the cases. A 1:1 ratio of controls to cases was employed for participants with reference dates from 2004–2007 and was then switched to 0.7:1 for those with reference dates from 2008–2010 as a result of additional funding that was obtained to increase the number of cases enrolled during the latter years of our data collection effort. A total of 90,488 random telephone numbers were pursued with multiple rounds of contacts attempted as needed. 66,844 numbers were nonworking, business, cellular, paging, dedicated facsimile, or data line numbers. 3,570 numbers were never answered, and thus their residential status could not be determined. Prior studies suggest that only about 20% of such numbers are indeed residential.(9
) Of the 20,074 residential or presumed residential numbers, 14,130 were successfully screened for eligibility. Of the remainder, 3,105 were answering machines, 2,351 reached a respondent who refused to answer the screening questions, and for 488 there were language or other communication barriers. Of the 1,489 eligible controls identified, 943 (63%) were interviewed.
For this analysis, 24 controls and 28 cases missing data on use of injectable contraceptives were excluded. Thus, our final analytic sample size consisted of 919 controls and 1,028 cases.
The study protocol was approved by the Fred Hutchinson Cancer Research Center Institutional Review Board, and written informed consent was obtained from all study subjects. Cases and controls were interviewed in-person and asked about their reproductive history, body size, medical history, and family history of cancer. Additionally, detailed histories of all episodes of hormonal contraceptive use, including beginning and ending dates, brand, dose, route of administration, and pattern of use (number of days per month) were obtained. Our questioning was limited to exposures that occurred before each participant’s reference date (month and year). The reference date used for each woman with breast cancer was her diagnosis date. As described above, controls were frequency matched to cases on reference year. The reference months assigned to controls reflected the distribution of reference months among the cases.
The primary referent category consisted of women who never used any type of injectable hormonal contraceptive. We also conducted sensitivity analyses considering alternative reference categories. These included defining women who never used any type of hormonal contraceptive as the reference category, and conducting analyses limited to ever users of hormonal contraceptives where the reference category consisted of ever users of a non-injectable hormonal contraceptive. Since even a single DMPA shot has been shown to result in MPA measurable in serum for as long as 7.5 to 9 months after injection, ever users of DMPA were defined as women who ever received even a single DMPA shot.(10
) Given these pharmacokinetics, recent DMPA use was defined as having received one or more DMPA shots within five years of reference date similar to how recency of hormone use has been defined by the Collaborative Group on Hormonal Factors.(12
) Duration of use was calculated by attributing three months of “use” (exposure) for each DMPA shot a woman received as the recommended prescribed regimen is shots that are administered every three months. Analyses focus primarily on recent DMPA use for at least 12 months because in pooled data from 54 epidemiological studies conducted world-wide the positive relationship between oral contraceptive use and breast cancer risk was only observed among recent users for at least 12 months duration.(13
We used unconditional logistic regression to calculate odds ratios (ORs) and their associated 95% confidence intervals (CIs) to compare breast cases to controls.(14
) All analyses were conducted using Stata/SE version 11·2 (StataCorp LP, College Station, TX). All models were adjusted for age (five year categories) and reference year (continuous) since controls were matched to cases on these factors and additionally adjusted for first degree family history of breast cancer (no / yes / missing), body mass index one year prior to reference date [<25.0 kg/m2
/ (25.0–29.9 kg/m2
/missing (based on WHO categories)], number of full-term pregnancies (0/1–3/≥3/missing), duration of oral contraceptive (OC) use (never/<5 years/5–9.9 years / ≥10 years / missing), and screening mammography (ever / never). These latter five covariates were selected a priori
as potential confounders. Other variables evaluated as potential confounders included: education, income, race/ethnicity (based on self-report), and age at first live birth. None of these potential confounders changed our risk estimates by more than 10% and there was no statistically significant change (p<0.05) in the fit of the model with the addition of any of these potential confounders. Thus none were added to our final statistical models.