Fracture Risk in Older, Long-term Survivors of Early-Stage Breast Cancer

Abstract

Background

Breast cancer survivorship has increased and the association of long-term bone health with breast cancer and its treatment is unclear. This issue is important for older women who face an increased risk of osteoporotic fracture-related morbidity and mortality.

Objectives

To examine the effect of breast cancer and its treatment on fracture risk in older breast cancer survivors and non-breast cancer comparisons.

Design

A 10-year prospective cohort study beginning 5 years after a diagnosis of breast cancer for survivors and match date for comparison women.

Setting

The study was conducted within six integrated health care systems.

Participants

Women 65 years and older who were alive and recurrence free 5 years after a diagnosis of early-stage breast cancer and matched on age, study site, and enrollment year to a non-breast cancer comparison cohort.

Measurements

Cox proportional hazards models were used to estimate the association between fracture risk and survivor-comparison status, adjusting for drugs and risk factors associated with bone health. A subanalysis was conducted to evaluate the association between tamoxifen exposure and fracture risk.

Results

Our analysis included 1,286 survivors and 1,286 comparison women with a mean age of 77.7 years in both groups. Survivor and comparison women were predominantly white and non-Hispanic, 81.6% and 85.2%, respectively. We observed no difference in fracture rates between groups (HR=1.1; 95%CI 0.9, 1.3). The protective effect of tamoxifen was not statistically significant (HR=0.9; 95%CI 0.6, 1.2).

Conclusion

Our findings suggest long-term survivors of early stage breast cancer diagnosed at age 65 and older are not at increased risk of osteoporotic fractures compared to age-matched non-breast cancer women. There appears to be no long-term protection from fractures following tamoxifen use.

INTRODUCTION

The number of breast cancer survivors has increased in the US since early 1970.(1) Currently, there are 3 million survivors and that number is estimated to increase.(1, 2) Breast cancer mortality increases with age;(3) however, women are often diagnosed after age 65 (40%) and at a localized stage (60%),(4) therefore treatment effects among survivors remains an important focus.(5) Tamoxifen has been, and remains a clinically important therapy for postmenopausal, hormone-positive breast cancer, especially in women unable to tolerate aromatase inhibitors (AIs).(6) Recent data from the Adjuvant tamoxifen: longer against shorter (ATLAS) trial show 10 years of tamoxifen further reduced breast cancer recurrence 25% and mortality nearly 30%.(7) Thus, tamoxifen is likely to remain an important therapeutic option.

The association of long-term bone health with breast cancer and its treatment is unclear. This issue is of particular importance for older women already at an increased risk of osteoporotic fracture-related morbidity and mortality.(8) Most notably, the incidence of major osteoporotic fractures per 1,000 women increases from 11.9 at 65 years to 40.4 after 85 years(9), and older women have been shown to experience functional decline following even a wrist fracture.(10) There is a paucity of information surrounding the impact of breast cancer and its treatment beyond diagnosis and early treatment. Kanis, et al. showed the prevalence of vertebral fractures in women with nonmetastatic breast cancer was similar at breast cancer diagnosis with controls, but fracture rates among breast cancer patients increased significantly over the following 3 years.(11) Nonetheless, the bone health of long-term (more than 5 years post diagnosis), early-stage breast cancer survivors is not well understood, and studies assessing the risk of hip fracture among survivors have demonstrated conflicting results.(12, 13) Lamont, et al(12) observed a decreased risk of fractures among women in the SEER-Medicare population, whereas Chen, et al(13) found a higher risk of fractures in breast cancer survivors from the Women’s Health Initiative. However, a gap in evidence surrounding the baseline fracture risk among older breast cancer survivors compared with non-breast cancer controls beyond 5 years of survival remains. Additionally, data suggests tamoxifen decreases fracture risk but the duration of reduced risk appears limited to the treatment period and the year following.(14)

The current study was conducted to provide an improved understanding of the effects of early-stage breast cancer and its treatment on bone health and fractures in older long-term survivors to inform survivorship care among these women. This study also provides data on fracture risk in non-breast cancer comparisons, providing fracture rates in older women adjusted for medications affecting bone health and other relevant risk factors.

METHODS

Setting

We conducted a prospective cohort study within six integrated health care systems participating in the National Cancer Institute (NCI)-funded Cancer Research Network (CRN). Study sites included: Group Health, Washington; Reliant Medical Group, Massachusetts; HealthPartners, Minnesota; Henry Ford Health System, Michigan; Kaiser Permanente Southern California; and Lovelace Health System, New Mexico. Institutional Review Board (IRB) approval, including a waiver of consent, was obtained at all study sites.

Population

Breast cancer survivors were identified in The Breast Cancer Treatment Effectiveness in Older Women (BOW I) study, which sought to evaluate treatment effectiveness in women 65 years and older diagnosed with early-stage breast cancer. Study methods relevant to BOW I have been described previously.(15) A follow-up study, Long-term Survivorship in Older Women with Early-stage Breast Cancer (BOW II), was designed to assess the late effects of cancer treatment. BOW II included 5-year breast cancer survivors (n=1,361) from the BOW I study cohort, and a comparison cohort without a breast cancer history (n=1,361). Breast cancer survivors were enrolled based on diagnosis date and comparisons were matched to survivors on age, study site and enrollment year with the survivor’s diagnosis date serving as the match date. Comparison women could not have evidence of a clinically active malignancy within the five years prior to study entry and were selected at random from a pool of eligible matched women at each site.

To eliminate the inclusion of subjects with new exposures to tamoxifen or aromatase inhibitors, a total of 75 breast cancer survivors, and their matched comparison subjects, were excluded from the current analysis due to breast cancer recurrence in the 5 years preceding the study period. Breast cancer survivors diagnosed with a recurrence or new breast cancer and comparison women diagnosed with breast cancer during follow-up were censored at the date of diagnosis. Women were followed beginning at year 5 and continued for up to 10 years.

Data collection

As previously described,(16) we used an electronic data collection system that incorporated preloaded electronic data from cancer registry (4 sites) and administrative databases (all sites) with manual medical record review to verify or supplement electronic data and to provide information for which electronic data were not available. The index date was defined as the date of breast cancer diagnosis for the survivor cohort and the match date for the comparison cohort. Data were collected beginning 5 years following diagnosis for breast cancer survivors and the match date for comparison women and continued for 10 years, i.e., through year 15 following diagnosis for breast cancer survivors and match date for comparisons. Data on age, race/ethnicity, body mass index (BMI), smoking status, Charlson comorbidities(17), hypertension, diabetes, rheumatoid arthritis, fractures of the hip, wrist, and vertebrae, and incident and recurrent breast and other cancers were collected. Each new occurrence of fracture of the hip, wrist and vertebrae substantiated by imaging was collected by medical record review. Relevant search terms were provided to data abstractors. Missing ethnicity was assigned by a validated adaptation of the Generally Useful Ethnic Search System computer program based on the subject’s surname.(18, 19) Comorbidity scores were calculated based on the Charlson Comorbidity Index (CCI) and updated biennially throughout the BOW II study period. Data specific to the breast cancer survivor cohort collected during the period prior to 5-years post diagnosis was collected during the BOW I study. These variables included Charlson comorbidities, BMI, smoking status, diagnosis year, tumor stage, and treatment variables such as receipt and length of tamoxifen treatment.

Automated filled pharmacy records were accessed to determine oral prescription medication utilization that either increase or decrease the risk of osteoporosis, including the selective estrogen receptor modulators (SERMs), AIs, bisphosphonates, estrogen, progestin, and corticosteroids (Supplemental Table 1). Generic Product Identifier (GPI) (Master Drug Data Base v2.0, Medi-Span, Indianapolis, IN) and National Drug Codes (NDC) were used to identify pharmacy claims for survivors and comparison women.

Statistical analyses

The primary study outcome of interest was first hip, wrist, or vertebral fracture. Women were censored at the earliest of these events: breast cancer diagnosis (comparison women), breast cancer recurrence including contralateral breast cancer (survivors), other cancer diagnosis, termination of health plan membership, date of death, or completion of 10 years of follow up.

Frequency distributions, chi-square tests, and Student’s t-tests were used to compare demographic characteristics of survivor and comparison groups. Kaplan-Meier estimates were used to compare the cumulative fracture rates over time. A series of Cox proportional hazards models were used to estimate fracture risk with survivor-comparison status. The base model was adjusted for age and study site, followed by four models adjusting for additional covariates: 1) race and BMI; 2) smoking status and comorbid conditions (CCI, hypertension, osteoporosis, arthritis); 3) prescription drugs associated with an impact on bone health (Supplemental Table 1); and 4) all covariates combined.

A subanalysis was conducted on the 1,286 breast cancer survivors who received tamoxifen (n=852) a survivors who did not receive tamoxifen (n=434) during the BOW I study period to assess the association of hip, wrist, and vertebral fractures and tamoxifen exposure. The analysis incorporated a second series of models. The base model, adjusted for study site and age, was followed by five models adjusting for 1) race and BMI; 2) tumor and treatment variables at diagnosis (i.e., surgery type, stage, estrogen, and progesterone receptor status, lymph node status, histology, and receipt of chemotherapy); 3) smoking status and comorbid conditions (i.e., CCI, hypertension, osteoporosis, and arthritis); 4) drug exposures associated with an impact on bone health (Supplemental Table 1); and 5) all covariates combined. All analyses were performed in SAS Version 9.2 (SAS Institute Inc., Cary, NC).

RESULTS

The study included 2,572 subjects, 1,286 survivors, and 1,286 comparison women with a median follow up of 6.0 and 6.5 years for survivors and comparison women, respectively. The mean age was 77.7 years in both groups (Table 1). Subjects were predominantly white and non-Hispanic (82% and 85% for survivors and comparison women, respectively). BMI measures and smoking status were similar in both groups. Most survivors and comparison women were identified as never-smokers or nonsmokers (64% and 62%, respectively). More than half (56% and 57% survivors and comparison women, respectively) had a CCI score equal to 0. Osteoporosis was diagnosed in 35% and 34% of survivors and comparison women, respectively. Survivors were predominantly diagnosed with stage I breast cancer (61%), 30% were diagnosed with stage IIa, and 9% with stage IIb disease.

Table 1.

Demographic characteristics of older, long-term survivors of early-stage breast cancer and age-matched comparison women

Characteristic n=	Survivor cohort n=1,286	%	Comparison women n=1,286	%
Age (years)
70-74	473	37	473	37
75-79	395	31	395	31
≥80	418	32	418	32
Race/ethnicity
White, non-Hispanic	1050	82	1095	85
African-American	132	10	118	9.2
Asian/Pacific Islander	33	2.6	26	2.0
Hispanic	71	5.5	47	3.6
Smoking status
Never/nonsmoker	824	64	802	62
Current	74	5.8	87	6.8
Previous	321	25	323	25
No mention	67	5.2	74	5.7
Body mass index (kg/m2)*
<18.5	34	2.7	45	3.5
18.5-24.9	354	28	381	30
25-29.9	356	28	357	28
≥30	282	22	231	18
Missing	260	20	272	21
Charlson Comorbidity Index
0	716	56	726	57
1−2	463	36	465	36
3−4	86	6.7	79	6.1
≥5	21	1.6	19	1.5
Other relevant diagnoses
Type I/II diabetes	194	15	168	13
Rheumatoid, other arthritis (714, 714.2)	66	5.1	75	5.8
Osteoporosis (733-733.09)	444	35	435	34
breast cancer recurrence (survivors)/ new breast cancer diagnosis (comparisons)	161	13	35	2.7

^*kilograms/square meter

Drug exposures for relevant osteoporotic treatments (i.e., bisphosphonates, calcitonin) were similar between survivors and comparison women (16% vs. 14% and 3.7% vs. 3.4%, respectively). Oral corticosteroids were also used with similar frequency between groups (23% vs. 23%). Hormone replacement therapy was more commonly used by the comparison women, whereas tamoxifen was used almost exclusively by survivors. We identified 68 (5%) survivors who received chemotherapy and tamoxifen and 43 (3%) who received chemotherapy alone. Aromatase inhibitors were used in 98 (6%) of survivors and 14 (0.8%) of matched comparisons. Raloxifene was used by 22 (1%) survivors and 13 (0.7%) matched comparisons.

We evaluated the incidence of first fracture for survivors and comparison women and found similar patterns over time. Incident rates of fractures were similar for survivors and comparison women (Table 2). Fractures of the hip, wrist, and vertebrae combined occurred at rates of 35 and 33/1,000 person-years for survivors and comparison women, respectively (P=0.51). Roughly 10% of women in each group experienced a fracture of the hip, wrist, or vertebrae.

Table 2.

Rates of incident osteoporotic fractures among older, long-term survivors of early-stage breast cancer and age-matched comparison women

Fractures	Survivors	IR*/1000 person years	Comparison	IR*/1000 person years
Person Time (Years)	8724		8874
Hip	117	14	130	15
Wrist	62	7.1	54	6.1
Vertebrae	122	14	106	12
Hip/Wrist/Vertebrae	301	35	290	33
Hip and Wrist Only	179	21	184	21

IR = Incidence Rate

Survivors were no more likely to experience a hip, wrist, or vertebral fracture than comparison women, regardless of covariate model (Table 3). The base model, adjusted for study site and age, showed no difference in all fractures combined (hip, wrist and vertebrae) (HR=1.0; 95% CI 0.8, 1.2) as well as for hip and wrist fractures only (HR=1.0; 95% CI 0.8, 1.3). The addition of variables including race, BMI, smoking status, comorbid conditions, or relevant drug exposures (Supplemental Table 1) did not alter fracture risk. Likewise, the inclusion of all covariates combined, showed no statistically significant differences in fracture rate (HR=1.1; 95% CI 0.9, 1.3).

Table 3.

Cox proportional hazards models of osteoporotic fracture risk for older, long-term survivors of early-stage breast cancer versus age-matched comparison women

Covariates	All fractures		Hip and wrist fractures
	HR*	95% CI	HR*	95% CI
Model 1: Study site, age	1.0	0.8, 1.2	1.0	0.8, 1.3
Model 2: Study site, age plus race and body mass index	1.0	0.8, 1.2	1.0	0.8, 1.3
Model 3: Study site, age plus smoking status and comorbid conditions	1.0	0.8,1.2	1.0	0.8, 1.3
Model 4: Study site, age plus drug exposure**	1.1	0.9, 1.3	1.1	0.9, 1.4
Model 5: Study site, age plus all covariates: combined	1.1	0.9, 1.3	1.1	0.9, 1.4

^*HR = hazard ratio.

^**Drug exposure detail in Supplemental Table 1

A subanalysis of survivors who received tamoxifen (n=852) compared with those who did not (n=434), with an average age of 77 (SD 5.9) and 77 (SD 6.3) years, respectively, was conducted to evaluate the impact of tamoxifen exposure on bone health. Both cohorts were similar with respect to race/ethnicity, smoking status, relevant drug exposures, and diagnoses associated with bone health (Table 4). Adjusted models were used to compare survivors who received tamoxifen with those who had not (Table 5). Adjusted models were used to compare survivors who received tamoxifen as part of their initial treatment (67%) with those who had not received tamoxifen (Table 5). No statistically significant differences in fracture risk were found among any of the models. The model containing all covariates combined showed a very small but non-significant protective effect for tamoxifen and was associated with a wide confidence interval (HR=0.9; 95% CI 0.6, 1.2).

Table 4.

Demographic characteristics of older, long-term survivors of early-stage breast cancer who did and did not receive tamoxifen

Characteristic n=	Total n=1286	%	Tamoxifen n=852	%	No Tamoxifen n=434	%	P value
Age (years)
70–74	473	37	309	36	164	38	0.8604
75–79	395	31	263	31	132	30
≥80	418	33	280	33	138	32
Race/Ethnicity
White, non-hispanic	1050	82	687	81	363	84	0.0615
African American	132	10	84	10	48	11
Asian/Pacific Islander	33	3	27	3	6	1
Hispanic	71	6	54	6	17	4
Smoking Status
Never/non-smoker	824	64	549	64	275	63	0.9529
Current	74	6	50	6	24	6
Previous	321	25	210	25	111	26
No mention	67	5	43	5	24	6
Body Mass Index (BMI)
<18.5	34	3	22	3	12	3	0.0004
18.5–24.9	354	28	243	29	111	26
25–29.9	356	28	246	29	110	26
≥30	282	22	199	23	83	19
Missing	260	20	142	17	118	27
Charlson Comorbidity Index
0	716	56	462	54	254	59	0.0073
1–2	463	36	326	38	137	32
3–4	86	7	56	7	30	7
≥5	21	2	8	1	13	3
Other Diagnoses
Diabetes	194	15	130	15	64	15	0.8017
Arthritis	66	5	38	4	28	6	0.1259
Osteoporosis	444	35	293	34	151	35	0.8857
Breast cancer recurrence	161	13	121	14	40	9	0.0106

Table 5.

Cox proportional hazards models of osteoporotic fracture risk for older, long-term survivors of early-stage breast cancer exposed to tamoxifen versus survivors not exposed to tamoxifen.

	All fractures		Hip and wrist fractures
Covariates	HR*	95% CI	HR*	95% CI
Model 1: Study site, age	0.9	0.7, 1.3	0.9	0.7, 1.3
Model 2: Study site, age, race, body mass index	0.9	0.7, 1.2	0.9	0.7, 1.3
Model 3: Study site, age, tumor and treatment variables	0.9	0.6, 1.2	0.9	0.6, 1.3
Model 4: Study site, age, smoking status, comorbid conditions	1.0	0.7, 1.3	1.0	0.7, 1.4
Model 5: Study site, age, drug exposure**	0.9	0.6, 1.2	0.9	0.6, 1.3
Model 6: All covariates: Study site, age, race, body mass index, tumor and treatment variables, smoking status, comorbid conditions, and drug exposure	0.9	0.6, 1.2	0.9	0.6, 1.4

^*HR = hazard ratio.

^**Drug exposure detail in Supplemental Table 1

DISCUSSION

We found that older women who survive early-stage breast cancer beyond 5 years were no more likely to experience a fracture of the hip, wrist, or vertebrae than matched comparison women. Adjusted models controlling for study site, subject age, smoking status, race, BMI, diabetes, osteoporosis, arthritis, and relevant drug exposures showed no increase in risk for osteoporotic fractures among the breast cancer survivor cohort relative to matched comparison women. This study focused on assessing osteoporotic fracture risk following 5 years of survival from a breast cancer diagnosis and did not identify fractures during that time period. Controlling for these events would have biased results by removing that portion of the breast cancer/fracture association captured by earlier fractures. The follow up period beginning after 5 years of survival and comparison match was chosen to better understand the long-term fracture risk at the point of care where follow-up often ceases and changes from oncology to primary care.(20) These data suggest that older, long-term, early-stage breast cancer survivors are at similar risk of osteoporosis and fractures as otherwise healthy women of similar age and should not be evaluated for the risks of osteoporosis any differently than the general population. This study improves our understanding of fracture risk in current, long-term early stage breast cancer survivors and provides important baseline data for comparison of fracture risk with exposure to newer agents such as the AIs. Likewise, the subanalysis focused specifically on tamoxifen use among the breast cancer survivors found a very small, but non-statistically significant, protective effect.

With relative survival from localized breast cancer estimated to be 99% at 5 years and the majority of breast cancers (60%) diagnosed at an early stage, some older breast cancer survivors can expect to live many additional years.(4) Tamoxifen remains a widely accepted treatment option for postmenopausal women who cannot tolerate AI therapy and offers a cost-effective alternative to the AIs. Therefore, understanding the long-term osteoporotic fracture risk is critical to identifying how to best care for these women. Previous studies have demonstrated increased fracture rates among breast cancer patients after diagnosis(11), but there were conflicting results surrounding fracture risk beyond 5 years of follow up.(12, 13) Using data from the National Cancer Institute’s Surveillance Epidemiology and End Results (SEER)-Medicare linked database, Lamont, et al. found a decreased risk of hospitalization due to hip fractures among survivors of early-stage breast cancer vs. comparison women (RR=0.6; 95% CI 0.4, 0.9).(12) Conversely, using data from the Women’s Health Initiative, Chen, et al. found that postmenopausal breast cancer survivors are at increased risk for all fracture types (HR=1.3; 95% CI 1.2, 1.4). However, the fracture risk among the subset of hip fractures was not statistically significant (HR=0.9; 95% CI 0.6, 1.3).(13) In contrast to Lamont et al and Chen et al, we found no difference in fracture risk among long-term survivors and their matched comparisons. Differences in outcomes between the current study and those conducted by Lamont or Chen could be due, in part, to differences in study populations and the methods used to identify fractures. Lamont et al. evaluated survivors of early stage (0, 1, or 2) breast cancer aged 55–64 years from the SEER-Medicare linked database including SEER tumor registry data and used hospitalizations to identify hip fractures only. Chen et al evaluated women with any history of breast cancer who were enrolled in the Women’s Health Initiative and used self-reports through patient questionnaires to identify fractures of the hip, wrist or vertebrae. We included women from six health plans, aged 65 or older and diagnosed with early stage breast cancer (0, 1 or 2) for whom hip, wrist or vertebral fractures were identified through comprehensive medical record review. Lamont, et al. would not have identified fractures unless a hospitalization occurred and Chen, et al identified fractures only through patient self-report. The current study utilized comprehensive medical record review to obtain fracture data.

The National Surgical Adjuvant Breast and Bowel Project Breast Cancer Prevention Trial (P-1) demonstrated tamoxifen reduces fracture risk in women at increased risk for breast cancer.(21) Women were randomized to tamoxifen or placebo for 5 years. Most fractures (89%) occurred in women aged 50 years or older and tamoxifen treatment was associated with a 29% reduction in fractures for that age group (RR = 0.71, 95% CI = 0.52 to 0.97) based on 7-years of follow up. A direct comparison to our study is difficult based on the study populations (women at increased risk for breast cancer compared with women with early stage breast cancer matched to non-breast cancer comparisons), the age-specific analyses, and follow-up time (7 years from randomization vs. 5-years post Index Date through 15 years) but the data suggest a potential beneficial effect of tamoxifen. Likewise, a study by Cooke et al demonstrated that tamoxifen is associated with a reduction in fracture risk among women aged 50 years or older with a fracture by ICD-9 code and medical claim for fracture reduction or fixation but the risk reduction was not extended to those with use within 1 year (OR=1.0; 95% CI 0.7, 1.4,) or past use (>1 year) (OR=1.1; 95% CI 0.8, 1.4).(14) The present substudy compared early stage 5-year survivors who received tamoxifen with those who did not through 15 years post diagnosis and did not find that tamoxifen provided a substantial protective effect against fractures. Our estimate was measured by medical record review and adjusted for exposures to oral prescription medications and disease states associated with an impact on bone health. Our findings suggest the ability of tamoxifen to preserve bone mineral density in postmenopausal women may not last beyond the treatment period, and may leave long-term survivors who are exposed to tamoxifen monotherapy at similar risk of osteoporotic fractures as age-matched comparison women.

Likewise, a recent historical cohort analysis of women with invasive breast cancer diagnosed between 1990 and 1999 showed no increased risk or protection from fracture as a result of a breast cancer diagnosis.(22) Our analysis demonstrates consistent results.

The timing of the diagnosis and treatment of patients enrolled in the present study resulted in tamoxifen being the primary hormone agent used for those who received hormone therapy. Five years of tamoxifen was, and still remains, a treatment option for postmenopausal, early stage breast cancer patients. AIs were not a primary therapeutic option until 2004;(23) and many older early stage survivors who received 5 years of tamoxifen remain in the US healthcare system. Likewise, some patients cannot tolerate the AIs or have significant cardiovascular risks making them a less attractive treatment option.(24–26) Davies, et al. found 10 years of tamoxifen is associated with a 25% reduction in breast cancer recurrence and nearly a 30% reduction in mortality among 6,846 women with ER-positive early stage breast cancer.(7) This provides important data regarding tamoxifen for whom the AIs are not suitable and demonstrates long-term tamoxifen therapy benefits outweigh the risks.(27) This data has been described as potentially practice-changing because there are no data to show that in postmenopausal women, AIs alone or in combination with tamoxifen are optimal compared with 10 years of tamoxifen. This may lead to more postmenopausal early breast cancer patients receiving 10 years of tamoxifen.(28)

This analysis informs our understanding of the baseline and long-term tamoxifen-related fracture risks and provides critical baseline data for future long-term outcome studies among patients who receive combined tamoxifen and aromatase inhibitors relative to the control population. AI use may impact fracture risk;(29, 30) however, the measure of this effect requires comprehensive baseline data on the fracture risk in older breast cancer survivors compared with non-breast cancer controls. This study provides that data with adjustments for other drugs that affect fracture risk and for other risk factors (e.g. age, smoking status, BMI, race/ethnicity, and comorbidities). Without the current data, the baseline risk for fractures that will enable the projection of risks and benefits of newer therapies is missing.

Prescription medications have been shown to have varying influence on bone health. In our analysis, we included oral drugs directly associated with breast cancer treatment (tamoxifen and AIs), osteoporosis (bisphosphonates, calcitonin, and raloxifene), menopause (estrogen and progesterone) and corticosteroids and found no difference in osteoporotic fracture risk among breast cancer survivors compared with matched control women. We identified 36 subjects exposed to AIs, 31 who received tamoxifen, 5 who did not, and 14 matched comparisons. Although we cannot explain AI exposure among the non-breast cancer matched comparison women, AI use was controlled for in the multivariate analyses. We believe the limited AI exposure identified among survivors occurred because of the timing of enrollment (1990–1994) and the availability of the aromatase inhibitors.

We censored women in both study groups who were diagnosed with pathologic fractures, a breast cancer recurrence (survivor cohort), or a new breast cancer (comparison cohort) during the study period to eliminate the exposure to medications that would impact fracture risk.

The strengths of our study include a large, geographically diverse population and the use of an age-matched comparison group. The baseline prevalence of risk factors for osteoporosis was similar in our survivor cohort and comparison group, including smoking status, BMI, CCI score, osteoporosis, diabetes, and arthritis. Additionally, our study incorporated administrative and claims data coupled with medical record review offering a rich dataset not previously included in similar analyses. Likewise, fracture data was obtained from chart review whereas prior studies have relied on either medical claims or patient recall for fracture occurrence. Last, our study included a long period of follow-up (10 years) beginning five years after diagnosis. A similar study with a follow-up period of this magnitude has not been previously conducted.

Study limitations include the lack of bone mineral density measures, fracture history in the 5 years preceding the study and the effect of injectable drugs such as the IV bisphosphonates on bone health. We were unable to ascertain which women in either group had a history of bone fractures, which could increase the risk for subsequent fractures. Our data are lacking several variables (prior fragility fractures, parenteral history of fracture, ever long-term use of glucocorticoids, daily alcohol consumption and bone mineral density). When FRAX variables are not available, any calculated fracture risk will be inaccurate based on the FRAX limitations.(31) Our survivor population, compared with an age-matched cohort, is similar based on BMI, smoking status, and diagnosis of rheumatoid arthritis. We believe that overall, both cohorts are at low fracture risk and it is unlikely that there are differences from other known causes of fractures between the groups. The FRAX score is based on hip bone mineral density and this could underestimate fracture risk if the impact is greatest on trabecular bone which is less prominent in the hip and more prominent in the vertebrae.(32) Additionally, it is possible patients received medical care at an external facility; however, patients included in this study were either enrolled and treated in completely integrated health care systems or were treated by provider groups with risk contracts with the associated insurer. We believe it is likely that all fractures were captured. Additionally, we did not have access to patient-reported data such as calcium intake, diet, physical activity, or adherence analyses for oral medication exposures in either group; however, the fractures rates in the two cohorts are consistent with other studies in this age group.(33, 34)

The study data is applicable to older, early stage (0, I or 2) breast cancer survivors who received adjuvant therapy with five years of tamoxifen and age-matched non-breast cancer comparison women and suggest that older breast cancer survivors beyond 5 years from diagnosis are at similar risk of osteoporotic fractures to that of age- and site-matched comparison women. Also, our data suggest that tamoxifen affords minimal or no long-term protection. Prior research suggests the effects of estrogen exposure decrease with time. Women who have stopped estrogen therapy for more than 10 years appear to have almost the same bone density and fracture risk as women who never received estrogen therapy.(35) Our study is consistent with those findings. In both breast cancer survivors and the comparison cohort, about 10% of women suffered a fracture over ten years of follow-up. These results, coupled with the high rate of long-term survival for early-stage breast cancer survivors, indicates a critical need for survivors to receive assessment and preventive care for fracture risks similar to those of older women in the general population. Thus, the incorporation of an osteoporosis screening and monitoring program in older, long-term survivors of early stage breast cancer is warranted.