Bisphosphonate Use and Risk of Recurrence, Second Primary Breast Cancer, and Breast Cancer Mortality in a Population-Based Cohort of Breast Cancer Patients

The efficacy of bisphosphonates in preventing age-related bone loss and reducing fracture risk in women with postmenopausal osteoporosis is well established (1–5). Preclinical data suggest that bisphosphonates may not only have positive effects on bone density and fracture risk but may also have potentially beneficial effects on the evolution of breast cancer. Bisphosphonates inhibit tumor growth and metastasis through inhibition of angiogenesis and tumor cell invasion, induction of apoptosis, and immunomodulatory effects (6–11). Bisphosphonates may also alter the microenvironment in the bone, creating a less hospitable environment for circulating tumor cell deposits (12, 13).

Clinical trials of bisphosphonate therapy for reduction in disease recurrence in women with early-stage breast cancer have had conflicting results. The ABCSG-12 trial suggested an improvement in disease-free survival with the addition of intravenous zoledronic acid to endocrine therapy in premenopausal women receiving ovarian suppression (14, 15). Two additional large randomized trials, the NSABP B34 study (16), which evaluated the addition of oral clodronate to standard adjuvant therapy, and the AZURE trial (17, 18), which examined an intensive dosing schedule of zoledronic acid, had overall negative results, but post hoc subgroup analyses revealed benefits in postmenopausal women. A recent meta-analysis of individual level data from over 18,000 breast cancer patients treated in randomized clinical trials of bisphosphonates versus control found statistically significant reductions in the risks of distant recurrence, bone recurrence, and breast cancer mortality among postmenopausal women, but no discernible effect in premenopausal women (19–21).

Several observational studies have examined the role of bisphosphonates not intended as cancer therapy in women with breast cancer. In a large cohort of primarily postmenopausal women with stage 0–III breast cancer treated with tamoxifen (median follow-up of five years), Kremer and colleagues (22) found significant decreases in the risk of bone metastases (OR = 0.50) and all-cause mortality (OR = 0.43) with bisphosphonate use. In the BRITE study, bisphosphonate use was associated with a reduced risk of second primary contralateral breast cancer, with stronger reductions seen with current use and longer duration of use (23). Conversely, an analysis of oral bisphosphonate use among breast cancer patients treated in the Kaiser Permanente health care system found no association with either recurrence or contralateral breast cancer (24).

Alendronate is the most commonly prescribed bisphosphonate for osteoporosis, but clinical trials have not assessed it in the context of preventing recurrences. In addition, the current clinical trial data have relatively short follow-up, and have been unable to assess the effect of BP use on later recurrences. Herein, we examine the association of bisphosphonate use for noncancer therapy and risk of adverse breast cancer outcomes in a cohort of women diagnosed with invasive breast cancer at ages 45 to 79. Our study included an evaluation by menopausal status, assessed patients with both estrogen receptor (ER)–positive and ER-negative disease, and had a median follow-up of 11.8 years.

The Quilt Study is a population-based prognostic cohort study of 2,337 women diagnosed with invasive breast cancer at ages 45 to 79 from 1993 to 1999. As reported previously, cases were ascertained through the Cancer Surveillance System, the population-based SEER registry covering western Washington State, and the cohort was formed by joining the cases recruited into three previous population-based studies of breast cancer, the CARE, PACE, and EMF studies (25). Briefly, data collection included in-person interviews, telephone interviews, and comprehensive medical record reviews. Information was collected on exposures before and after diagnosis, including reproductive history, exogenous hormone use, medical history, mammography, height, weight, smoking, alcohol, and family history; tumor features/markers; treatment history; and ongoing follow-up for recurrence and/or death. Menopausal status was defined as premenopausal, perimenopausal, or postmenopausal using data on last menstrual period, hormone replacement therapy, and oophorectomy/hysterectomy status collected at interview, as previously described (26). Because of small numbers, all analyses (after Table 1) combined pre- and perimenopausal women into one non-postmenopausal group. All participants provided written informed consent and our institutional review board approved the study protocol.

Of the 2,337 eligible breast cancer cases, we restricted to early-stage women by excluding those with distant (n = 53) or unknown (n = 3) stage at diagnosis. We further excluded those with incomplete medical record data on bisphosphonate use or recurrence (n = 466), and those with another cancer diagnosed at the same time as their breast cancer (n = 2), resulting in 1,813 women with invasive breast cancer available for analysis. Compared with excluded women, those included were significantly more likely to be older, postmenopausal, treated with adjuvant hormonal therapy, not treated with chemotherapy, more recently diagnosed, and to have received a recent screening mammogram.

Women were followed until the first of any of the following types of events: recurrence of breast cancer (any, locoregional, distant); new breast primary; other cancer; death; or until the end of follow-up. Other nonbreast subsequent primary cancers were treated as censored in analyses. Breast cancer–related deaths in the absence of a reported distant recurrence were deemed as having been preceded by a distant recurrence (n = 16), with the event date imputed using median time from distant recurrence to breast cancer–related death in those with such information available, or last date known to be distant recurrence–free, whichever was later. Events in the contralateral breast were considered new breast primaries. Median follow-up time in those without events was 11.8 years (range, 2.5–17.6) for recurrences and 15.1 years (range, 0.5–21.6) for mortality. Follow-up time in BP users and nonusers was similar, with more than 80% of those without events in each group being followed more than 10 years.

Information on bisphosphonate use was obtained from medical record reviews. When a drug appeared in the medical record, abstractors recorded the date, name and type, and whether the drug was started, stopped or being used at that time. For intravenous use of Reclast, we assumed 12 months exposure duration, and for intravenous Zometa and pamidronate, we assumed 6 months of exposure. We generated an electronic calendar of use of bisphosphonates for each month from diagnosis to first event or time of last follow-up. We assumed continued use after a “start” or “on” notation, until a “stop” or the last mention in the chart. This facilitated calculation of time-varying exposure variables for analysis. Women with no reported use (n = 1,458) and those with only one mention of use (n = 53) were classified as nonusers. Bisphosphonate use as treatment for bone or other metastases was excluded. In our analyses of breast cancer mortality, we restricted to use before first event, as our focus was on nontreatment-related bisphosphonate use. Prediagnosis BP use (n = 21 women) was ignored.

Cox proportional hazards regression methods were used to calculate HRs and 95% confidence intervals (CI) for the association of bisphosphonate use and the various time-to-event outcomes. Postdiagnosis bisphosphonate exposure, including duration and current (within 6 months) use, was calculated as of the time of each event in the cohort, with all women still event-free and under observation at that time included in each comparison. All analyses were adjusted a priori for age at diagnosis and source study (the original case–control studies from which cases were ascertained: CARE, PACE, EMF). Factors examined as potential confounders (10% change in HR) and effect modifiers (likelihood ratio test interaction P < 0.05) include diagnosis age, diagnosis year, source study, race, BMI at diagnosis, screening mammography, menopausal status at diagnosis, prediagnosis alcohol and smoking history, comorbid conditions, family history, diagnosis with osteoporosis or osteopenia, stage, grade, ER/PR status, and treatment (radiotherapy, chemotherapy, hormonal treatment, tamoxifen duration as a time-varying covariate). Proportional hazards assumptions were assessed using Schoenfeld residuals. Analyses were conducted in Stata version 14. A two-sided P value <0.05 is considered statistically significant.

Because the absence of mention of bisphosphonate use in medical records was our only indication of nonuse, we performed sensitivity analyses that excluded women with less complete chart reviews, specifically those who had no weight measurements and those with an indication of a gap in medical record review. Similar results were seen to those presented below.

Characteristics of the study cohort are presented in Table 1. The mean age at diagnosis was 64.2 years and 84% were postmenopausal at diagnosis. A majority of patients were ER or PR positive (86% and 77%, respectively); 67% received hormonal therapy, primarily tamoxifen. Seventeen percent of participants (n = 302) used a bisphosphonate after diagnosis. Of these, most (83%) had one or more years of use. The majority (96%) of bisphosphonate use was oral (90% used alendronate and 13% used risendronate); 4% of users took intravenous zoledronate. The mean duration of postdiagnosis bisphosphonate use was 41.2 (range, 2–145) months. Ninety percent of bisphosphonate users and 35% of nonusers had a recorded diagnosis of osteoporosis or osteopenia.

Bisphosphonate use was associated with a significantly decreased risk of any breast cancer event (locoregional/distant recurrence or second primary breast cancer; Table 2). Ever (vs. never) use of bisphosphonates was associated with a HR of 0.65 (95% CI, 0.47–0.90). The HR for each year of continuous use was 0.87 (0.78–0.96). Use for one or more years of use was associated with a 40% reduction in risk (95% CI, 0.41–0.88) with a suggestion of a trend toward lower risk with longer duration of use (P = 0.096). A slightly stronger effect was seen for current use of bisphosphonate versus former use (current use HR, 0.54; 95% CI, 0.34–0.86; former use HR, 0.80; 95% 0.51–1.24). An earlier start of bisphosphonates after diagnosis was associated with stronger reductions in risk of recurrence or second primary breast cancer. Specifically, those who began bisphosphonates within the first 3 years after diagnosis had a HR of 0.47 (95% CI, 0.26–0.86) for any breast cancer event whereas those who started later had a HR of 0.72 (95% CI, 0.45–1.16). Additional adjustment for other variables in Table 1 did not substantially alter these results, and subsequent analyses adjusted only for age and source study. Excluding the 12 women who ever used injectable bisphosphonates (7 of whom also used oral bisphosphonates) had no effect on risk estimates.

To assess whether the effect of bisphosphonates was limited to certain types of events, we assessed each type individually (Table 3). Although the numbers of patients per event type were smaller and power more constrained, a generally similar pattern of risk to that in the overall group was observed. Nonsignificant reductions in risk in relation to ever use of bisphosphonates use were observed for locoregional recurrence (ever use HR, 0.61; 95% CI, 0.26–1.43; 1+ years use HR, 0.7; 95% CI, 0.28–1.77), distant recurrence (ever use HR, 0.62; 95% CI, 0.36–1.06; 1+ years use HR, 0.51; 95% CI, 0.26–1.00) and second primary breast cancer (ever use HR, 0.75; 95% CI, 0.48–1.18; 1+ years use HR, 0.71; 95% CI, 0.43–1.17). Bisphosphonate use was associated with nonsignificant reduced risks for a first distant recurrence occurring in bone (ever use HR, 0.53; 95% CI, 0.24–1.15; 1+ years use HR, 0.41; 95% CI, 0.15–1.13) and outside of bone (ever use HR, 0.73; 95% CI, 0.35–1.52; 1+ years use HR, 0.63; 95% CI, 0.25–1.58).

As other studies have suggested that the protective effect of bisphosphonate use is limited to postmenopausal women, we assessed bisphosphonate effects by menopausal status (Table 4). Ever use of bisphosphonates was associated with reductions in risk of a breast cancer event in both groups, with small numbers but a greater magnitude of reduction in pre/perimenopausal women (ever use HR, 0.34; 95% CI, 0.12–0.94). Bisphosphonate use was associated with a significant 36% reduction in risk of any breast cancer event in ER-positive patients and a nonsignificant 54% reduction in risk among ER-negative patients (Table 5). While the number of pre-/perimenopausal ER-negative patients was small (n = 50), it is potentially striking that there were no distant recurrences in the 9 bisphosphonate users in this subgroup as compared with 12 recurrences (29%) in the 41 nonusers.

Our analysis of effect modification by factors in Table 1 yielded a suggestive pattern of larger reductions in risk of recurrence in relation to bisphosphonate use in women with poorer-prognosis tumors (regional stage, higher grade, ER-negative, and PR-negative disease). However, differences were only statistically significant for grade (well/moderate differentiated HR, 0.98, poor/undifferentiated HR, 0.34; P_interaction = 0.004). We also examined the effect of bisphosphonates by the interval events first occurred after initial diagnosis. Results were similar over time (for 1+ years use, HR for event in first 5 years, 0.62; 95% CI, 0.39–0.98; HR for event at 5–<10 years, 0.66; 95 % CI, 0.43–1.02; HR for event at 10+ years, 0.50; 95% CI, 0.25–0.97)

The use of a bisphosphonate for one year or more was associated with a significantly decreased risk of breast cancer mortality (HR, 0.40; 95% CI, 0.23–0.69; Table 6). Restricting the analysis to the first 7 years of use, for which our data were most robust, did not alter this result (HR, 0.40; 95% CI, 0.22–0.73).

In our study, postdiagnosis use of a bisphosphonate was associated with sizable (30%–50%) reduced risks of subsequent breast cancer events (locoregional recurrence, distant recurrence or second primary breast cancer) and breast cancer mortality. Reduced risks in relation to bisphosphonates were seen in both pre/peri and postmenopausal women, and in women with both ER-positive and ER-negative disease. In addition, we saw suggestion of a stronger effect with longer duration of bisphosphonate use and with an earlier start after diagnosis.

The data from clinical trials of BP administered as oncologic therapy have been inconsistent but suggest in the aggregate, as summarized in a recent patient-level meta-analysis (19), statistically significant 10% to 20% reductions in risk of recurrence and death among postmenopausal women and an absence of benefit in premenopausal women. However, taken together, results from our study and two other observational studies suggest a somewhat more nuanced picture, including larger magnitude (i.e., 30%–50%) reductions in risk of adverse breast cancer outcomes (22, 23) and beneficial effects in both postmenopausal and nonpostmenopausal or younger women (22). It is worth noting that a fourth observational study which combined data from two study sites found no overall benefits of bisphosphonates, although there was some suggestion of nonstatistically significant risk reductions at one of the two sites (24).

Follow-up length was different between the observational and clinical trial data. For example, the median follow-up in the aforementioned meta-analysis was 5.6-person years while the median follow-up in our study was twice as long (22, 23). It is also worth noting that the observational studies had the opportunity to assess several oral bisphosphonates (i.e., alendronate and risedronate) that have not been included in the clinical trials. A major difference between observational studies of bisphosphonate use and randomized clinical trials is the indication for use; in the observational studies, bisphosphonate prescription was completely or largely (depending on the study) targeted toward women with known loss of bone density whereas in the randomized trials, women received a bisphosphonate regardless of their bone density. We hypothesize that the beneficial effect of bisphosphonates on breast cancer recurrence and mortality risk is present in women with low bone density, regardless of menopausal status.

Another interesting finding in our study was a suggestive, though not statistically significant, reduced risk in women with ER-negative breast cancer and those with high-grade disease. The EBCTCG meta-analysis (19) and other studies (17, 18, 27) have shown similar findings. The possibility of a reduction in recurrence risk with bisphosphonate use in women with ER-negative breast cancer, for whom there is currently no treatment other than chemotherapy, is intriguing and warrants further study.

Notable features of bisphosphonates include high binding rates and long-term absorption in bone and relatively limited short-term systemic bioavailability (28), properties which would suggest that benefits might be more pronounced for bone versus other locations. Our results do suggest stronger risk reductions for bone metastases as first metastatic events but there were also sizable risk reductions for first metastases outside of bone and for breast cancer mortality. Further mechanistic research is needed to determine whether potential benefits for nonbone first metastases derive from direct effects of bisphosphonates or from disrupted pathways of metastatic spread that might otherwise start in bone (29).

A limitation of our study was the fact that we relied on medical record abstraction to obtain information on bisphosphonate, which may have been incomplete. We conducted sensitivity analyses, excluding women with less complete chart reviews and results were unchanged. We would not expect incomplete reviews to differ by recurrence status, and in fact might expect those with recurrences to be more likely to mention earlier use of medications at subsequent provider visits. Given our results were in the opposite direction, recall or misclassification bias seems less likely. Nonetheless, to the extent that the association between bisphosphonates and recurrence differed in included versus excluded women, our results could be biased or less generalizable. In addition, medical records do not provide information about actual medication use. Pharmacy records would have been helpful as a complementary data source for assessing adherence and more precisely measuring duration but were not available. There is also a concern that observational data are subject potentially to other sources of confounding or bias. For example, it is possible that patients who received bisphosphonates are at lower risk of breast cancer recurrence, or differ with regard to comorbidities. Although the numbers were small, our results did not substantially differ when we limited analyses to those with a diagnosis of osteopenia or osteoporosis, and covariate adjustment for such a diagnosis did not alter risk estimates. While BP users did appear to be less likely to have comorbidities, there was no attenuation of HRs following additional adjustment for the Charlson Comorbidity Index or smoking status. This population was mainly treated with tamoxifen, which could be considered both a potential strength and a weakness. On the positive side, results are not confounded by the expected bone loss caused by aromatase inhibitors and treatments were largely homogenous across the study period. On the negative side, as AIs are now considered the standard of care in postmenopausal women, the study conditions do not reflect the entirety of current practice.

Observational data can complement clinical trials in numerous ways. Our study has many strengths, and makes a unique contribution to the current literature on bisphosphonate use in breast cancer patients. First, the long median duration of follow-up allowed us to assess both early and late recurrences, which is of particularly crucial importance for patients with hormone receptor–positive disease. In this study, our data suggest that the effect of bisphosphonates does not diminish over time, and therefore protection against both early and late recurrences is observed.

Second, our study provides timely data on a bisphosphonate which is widely used for the treatment of osteoporosis but which has received little attention in relation to breast cancer progression. Both the recent ASCO/CCO Clinical Guideline on adjuvant bisphosphonates (30) and the EBCTCG meta-analysis (19) noted the complete lack of randomized clinical trial data on the effect of alendronate on breast cancer recurrence risk. Alendronate and risendronate belong to the class of aminobisphosphonates, which are thought to be more potent than non-nitrogen containing bisphosphonates such as clodronate, and similar in potency to intravenous zoledronic acid. Our data suggest that oral aminobisphosphonates are effective for reducing recurrence risk in women with breast cancer, and thus could represent a convenient and a more cost effective option for this population.

Notably and in contrast with the clinical trials, participants in this study were prescribed bisphosphonates due to bone loss. We hypothesize that the reason our results differ from the trials, both in stronger magnitudes of risk reduction and beneficial effects in postmenopausal and nonpostmenopausal women, is that the bone microenvironment created by low bone density results in a higher risk of recurrence, and this risk is abrogated by BP use. Thus, in randomized trials in which women were assigned to BP use regardless of bone density, the effects were more modest and only clear in postmenopausal women, in whom osteopenia and osteoporosis are most prevalent.

In this cohort study of women with breast cancer, we observed a sizable reduction in the risk of breast cancer events (distant recurrence, locoregional recurrence, or second primary breast cancer) among users of oral bisphosphonates. There was a stronger inverse association for use that started within 3 years of diagnosis versus later, and we observed reductions in risk for both early and late recurrences of breast cancer. Importantly, women who used bisphosphonates also had an improvement in breast cancer–specific survival compared with never users. Our results support the findings of randomized adjuvant studies, which showed a lower risk of recurrence in postmenopausal women receiving bisphosphonates, and suggest that premenopausal women with low bone density may benefit from this treatment as well. Finally, our data support the potential effectiveness of aminobisphosphonates such as alendronate and risendronate for the prevention of recurrence and mortality in women with breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: L.A. Korde, K.E. Malone

Development of methodology: L.A. Korde, K.E. Malone

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P.L. Porter, K.E. Malone

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.A. Korde, D.R. Doody, L. Hsu, K.E. Malone

Writing, review, and/or revision of the manuscript: L.A. Korde, D.R. Doody, P.L. Porter, K.E. Malone

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.R. Doody, K.E. Malone

Study supervision: K.E. Malone

This research was supported by NIH grants R01 CA098858 and R01 CA192438, and included salary support for the following authors: D.R. Doody, L. Hsu, P. Porter, and K.E. Malone.

The authors would like to thank the study participants for their invaluable contributions to this research, Cecilia O'Brien for excellent study coordination, and the data collection staff for their high-quality work.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.