Long-term use of calcium channel blocking drugs and breast cancer risk in a prospective cohort of US and Puerto Rican women

Statistical methods

We characterized use of calcium channel blockers and other antihypertensive drugs in the cohort using standard contingency table methods with chi-square tests. Univariate and multivariate logistic regression analyses were used to identify potential covariates associated with different levels of current and lifetime antihypertensive use across drug subclasses. All statistical analyses were conducted using SAS 9.3 (Cary, NC, USA).

To determine the association between calcium channel blocker use and breast cancer risk, we calculated hazard ratios (HRs) for strata of lifetime use and incident breast cancer using multivariate Cox proportional hazard regression. Calcium channel blocker use was categorized into five levels: (1) never used calcium channel blocker; (2) former user of calcium channel blocker; (3) current user with less than 5 years of use; (4) current user with 5–10 years of use; and (5) current user with 10 years or more of use. The same stratifications were used for use of any type of antihypertensive drug, as well as for the following classes of antihypertensive drugs: beta-blockers, angiotensin-converting enzyme (ACE) inhibitors, diuretics, angiotensin receptor blockers, and miscellaneous drugs. Calcium channel blockers were also stratified into dihydropyridines (amlodipine, felodipine, nifedipine SR, isradipine, nifedipine, nicardipine) and nondihydropyridines (verapamil, diltiazem, verapamil SR, diltiazem XR). Women who switched from one antihypertensive drug subclass to another subclass before the baseline survey were considered former users for their previous drug subclass, current users for their current subclass, and current users of antihypertensives in general.

Attained age serves as the time scale for the Cox proportional hazards regression models. Participants enter the analysis with their age at study enrollment, and accrue person-time until they either exit at the age of breast cancer diagnosis for cases or the age at their last completed health questionnaire. Participants who failed to respond to their most recent eligible health update were censored at the earliest date among death or the midpoint of the interval between the last completed health update and the end of the window of eligibility for responding to their first missed health update. We ran crude unadjusted Cox proportional hazards regression models as well as a multivariable model with adjustment for the following variables selected as a-priori potential confounders based on existing literature: baseline-measured BMI, race/ethnicity, smoking status, parity, menopause status, age at menarche, history of breast conditions, current drinking level, hormone replacement therapy use, and statin use. We also ran models excluding women who reported prophylactic mastectomy at baseline (N?=?234), or with adjustment for prophylactic tamoxifen and reloxifene use, but these factors had no material effect on our risk estimates and were excluded from the final model.

As an a-priori hypothesis, we examined risk associated with long-term use of calcium channel blocking drugs, and additionally examined use of antihypertensives in general and use of other subclasses of antihypertensive drugs. We focused on use of calcium channel blockers that lasted for 10 years or more, because this is the length of usage that has recently been associated with breast cancer risk. We also conducted analyses where long-term use was defined less strictly as 5 years or more of use to provide increased statistical power for risk detection. We examined risk associated with long-term calcium channel blocker use by breast tumor characteristics including tumor invasiveness at diagnosis and estrogen receptor status. To make our results as comparable as possible with previously conducted studies, we conducted sensitivity analyses limiting our analytical group to: women who were postmenopausal at study entry; and women who were aged 55 or older at study entry. Additionally, we conducted an active comparator sensitivity analysis in which we limited our analytical group to women who were current users of an antihypertensive drug: for at least 5 years; and for at least 10 years. This analysis allows for direct comparison of long-term calcium channel blocker users with women who were long-term users of another class of antihypertensive drug.