Regular and Low-dose Aspirin, Other Non-steroidal Anti-inflammatory Medications and Prospective Risk of HER2-defined Breast Cancer

The California Teachers Study

Christina A. Clarke; Alison J. Canchola; Lisa M. Moy; Susan L. Neuhausen; Nadia T. Chung; James V. Lacey Jr; Leslie Bernstein


Breast Cancer Res. 2017;19(52) 

In This Article


The California Teachers Study cohort (CTS) was established in 1995–1996 when 133,479 active and retired female teachers, administrators and other public school professionals were recruited through the California State Teachers Retirement System.[14] Participants completed a baseline questionnaire that collected information on family history of cancer and other conditions, menstrual and reproductive history, self-reported weight and height, living environment, diet, alcohol and tobacco use, physical activity history, and frequency and duration of prior use of certain medications including aspirin (but without detail on aspirin dose).

In 2005–2006, a 10-year follow-up questionnaire collected updated information on frequency of current use of aspirin, low-dose aspirin and other pain-relieving medications (see below), weight, alcohol use, menopausal status, use of hormone therapy (HT), and physical activity. Copies of questionnaires are available at The CTS is overseen by the Institutional Review Boards of the Cancer Prevention Institute of California, the California Health and Human Services Agency, the University of California, Irvine, the University of Southern California, and the City of Hope.

CTS participants are followed annually for changes of address, cancer diagnoses, hospitalizations, outpatient surgeries, emergency room visits, and death. Annual linkage with the California Cancer Registry (CCR) is used to identify incident cancer among cohort members. The CCR is a population-based cancer registry, which is anchored in state legislation that mandates reporting and is estimated to be over 99% complete.[15] Annual linkages with the Office of Statewide Health Planning and Development (OSHPD) allow us to identify details of each members' hospitalizations, outpatient surgeries, and emergency room visits. California and national mortality files are used to ascertain dates and causes of death.

Breast Cancer Ascertainment

Information on all incident breast cancers was obtained from the CCR, including pathologic and clinical features, which are abstracted directly from the medical record. HR status was based on estrogen receptor (ER) and progesterone receptor (PR) status as routinely reported on diagnostic pathology records. HER2 status was also based on pathology report review. Subtypes were defined as follows: HR-positive/HER2-negative was defined as ER-positive or PR-positive and HER2-negative; HR-positive/HER2-positive was defined as ER-positive or PR-positive and HER2-positive; HR-negative/HER2-positive was defined as ER-negative and PR-negative and HER2-positive; and triple-negative was defined as ER-negative, PR-negative, and HER2-negative.

Assessment of Pain-relieving Medications

On the 10-year follow-up questionnaire, women were asked whether they were currently taking any pain-relieving medications at least once a week, and if yes, the total number of tablets taken per week (1–2, 3–4, 5–6, 7–8, 9–10, 11–12, 13–14, 15–21, 22–28, or 29+ tablets per week). The medication choices included low-dose aspirin; aspirin or aspirin-containing product (Bayer, Bufferin, Excedrin); ibuprofen (Advil, Motrin); naproxen, ketoprofen or other non-steroidal (Aleve, Feldene, Indocin, Naprosyn, Orudis, Relafen); Cox-2 inhibitors (Celebrex, Vioxx); and acetaminophen (aspirin-free Excedrin, Tylenol, Tempra). Women were then asked if they had stopped regular use of any of these medications during the past 3 years, and if yes, why (by marking any of the following response categories that applied: "Condition improved", "Didn't work", "I had side effects", "I heard about side effects", "Drug no longer available", "Other"). The baseline questionnaire asked if aspirin (Anacin, Bufferin, Excedrin) and ibuprofen (Advil, Motrin, Nuprin) were taken regularly (at least once a week), total years taken and how many days per week taken (1–3 days/week, 4–6 days/week or daily).

Each medication type from the 10-year questionnaire was initially categorized as "Never in the past 3 years", "Former", "Current", or "Unknown" for that type of medication. Women were classified into the category "Never in the past 3 years" if they reported: (1) current use of 0 or <1 medication per week or left this question blank; (2) never took the medication regularly or did not stop regular use or left this question blank; and (3) did not give a reason for stopping. "Current" users were those who reported: (1) current use of ≥1 tablet(s) per week; (2) "Never took regularly or did not stop use" or left this question blank; and (3) did not give a reason for stopping. "Former" users were those who reported: (1) "Yes, I stopped regular use" or gave a reason for stopping; and (2) reported current use as 0 or <1 medication per week or left this question blank. All other women, including those whose answers were inconsistent or who left both questions blank for all medications, were classified as "Unknown".

For analysis, we: (1) focused on the five NSAIDs from the 10-year questionnaire, excluding acetaminophen; (2) grouped the women who were in the category of "Never in the past 3 years" for the medication of interest into women who reported (a) "No NSAIDs in the past 3 years," which was used as the reference group, (b) no use of that particular type of medication but had used one or more of the other four NSAIDs in the past 3 years, and (c) no use of that particular type of medication and unknown use of one or more of the other four NSAIDS, which was combined with the "Unknown" category; and (3) combined former and current users of 1–2 tablets per week into one category. For low-dose aspirin, we also examined "Daily" use, which was defined as those with current use of 7+ tablets per week (Table 2).

Study Population

For the present analyses, we excluded women sequentially who at baseline were not residing in California (n = 8867), had a prior history of breast cancer (n = 6216), or had unknown cancer history (n = 135). We also excluded women who prior to the 10-year follow-up questionnaire had died (n = 8654), had requested no further contact from the CTS (n = 926), had moved out of California for more than 4 months (n = 8296), had developed breast cancer (n = 4188), or had a bilateral mastectomy without a diagnosis of breast cancer (identified from hospital discharge data) (n = 18). Among the remaining 96,179 participants, 57,164 (59%) completed the 10-year follow-up questionnaire. During follow up (from the date a woman completed the 2005–2006 questionnaire and continuing through 31 December 2012), 1457 women were diagnosed with invasive breast cancer after completing the 10-year follow-up questionnaire.

Data Analysis

Follow-up time was calculated as the number of days between the date the 10-year follow-up questionnaire was completed and the first of the following events: a first diagnosis of breast cancer (International Classification of Diseases for Oncology-3 (ICD-O-3) site code C50) (n = 1457 with invasive cancer; n = 393 in situ cancer), death (n = 3538), a move (for >4 months) out of California (n = 2082), bilateral mastectomy (n = 5), or 31 December 2012 (n = 49,689).

Hazard rate ratios (HRR) and 95% CI were estimated using multivariable Cox proportional hazards regression models, with age (in days) as the time metric and stratification by age (in years) at the time of the 10-year follow-up questionnaire. Covariates were included based on their independent association with risk for a given outcome with a p value <0.05 in multivariable models.

Competing risk analysis was used to estimate risk of invasive breast cancer overall and by different receptor subtypes. The 157 women diagnosed with breast cancer during follow up who had missing information on HR or HER2 status were excluded from all models with breast cancer subtype as the outcome. We tested the proportional hazards assumption for each covariate in the model and for the main effect for the different outcomes using a likelihood ratio test of interaction with the time metric (continuous age) based on cross-product terms. We found only one violation of the proportional hazards assumption: alcohol consumption had a statistically significant interaction with time-dependent age for HR-negative/HER2-negative tumors; thus, this interaction was included in the model for that outcome.

We conducted multivariable analyses in which we assessed the impact of adjusting for history of hospitalization for myocardial infarction (between 1991 and the date the 10-year follow-up questionnaire was completed based on ICD-9 diagnostic codes 410.00–410.92 from OSHPD linkage; no, yes) and history of diabetes mellitus (from the 10-year follow-up questionnaire; no, yes, missing). These adjustments did not meaningfully change the HRR when evaluating the association of current use of low-dose aspirin with risk of breast cancer overall or with risk of any of the receptor subtypes. Hence, history of myocardial infarction and history of diabetes mellitus were not included in the final models presented here. We also examined models of low-dose aspirin stratified by body mass index (BMI) (<25 or ≥25 kg/m2) and tested the interaction using a likelihood ratio test and cross-product terms, excluding women with unknown NSAID use.

In a secondary analysis, we considered aspirin use at baseline (regular-dose and low-dose aspirin were not asked about separately) in conjunction with use of regular-dose and low-dose aspirin from the 10-year questionnaire, with no NSAID use reported at both baseline (including aspirin and ibuprofen) and the 10-year follow-up (including regular-dose and low-dose aspirin, ibuprofen, Cox-2 inhibitors, and other NSAIDs) as the reference group. This analysis was limited to outcomes of breast cancer and the HR-positive/HER2-negative subtype, as we had an insufficient number of diagnoses to examine these associations in the other subtypes.