Oral Bisphosphonate Use and Risk of Postmenopausal Endometrial Cancer

Abstract

Purpose

Bisphosphonates are common medications used for the treatment of osteoporosis and are also used to reduce metastases to bone in patients with cancer. Several studies, including the Women's Health Initiative (WHI), have found that use of bisphosphonates is associated with reduced risk of developing breast cancer, but less is known about associations with other common malignancies. This study was aimed at examining the effects of bisphosphonates on the risk of endometrial cancer.

Methods

We evaluated the relationship between use of oral bisphosphonates and endometrial cancer risk in a cohort of 89,918 postmenopausal women participating in the WHI. A detailed health interview was conducted at baseline, and bisphosphonate use was ascertained from an inventory of regularly used medications at baseline and over follow-up. All women had an intact uterus at the time of study entry.

Results

During a median follow-up of 12.5 years, 1,123 women were diagnosed with incident invasive endometrial cancer. Ever use of bisphosphonates was associated with reduced endometrial cancer risk (adjusted hazard ratio, 0.80; 95% CI, 0.64 to 1.00; P = .05), with no interactions observed with age, body mass index, or indication for use.

Conclusion

In this large prospective cohort of postmenopausal women, bisphosphonate use was associated with a statistically significant reduction in endometrial cancer risk.

INTRODUCTION

Bisphosphonates are widely used for the prevention and treatment of osteoporosis; in 2007, nearly 6 million patients received prescriptions for this indication.¹ Aminobisphosphonates, the most active of these drugs, also have cytostatic, proapoptotic, and antimetastatic properties² and have been inversely associated with breast cancer risk in some,^3–6 but not all,⁷ studies. Hormonally mediated endometrial cancer shares many risk factors with breast cancer,⁸ and previous studies have reported reduced risk of endometrial cancer in women with a history of fractures,^9,10 providing support for this thesis. A few small retrospective studies of bisphosphonate use and endometrial cancer have been published with inconsistent findings.^11–13 We examined this association in the well-annotated Women's Health Initiative (WHI), a prospective cohort involving four randomized clinical trials (WHI-CT) and a large observational study (WHI-OS). Use of oral bisphosphonates was ascertained for WHI participants at baseline and over the follow-up period.

METHODS

Study Population

WHI enrolled 161,808 postmenopausal women, age 50 to 79 years, from 40 clinical sites across the United States between 1993 and 1998. WHI-OS participants were generally similar to those of WHI-CT but were ineligible or unwilling to be included in a randomized trial. Details of the WHI recruitment, eligibility criteria, and protocols have been published elsewhere.^14–17 All participants provided informed consent. The institutional review boards at all WHI institutions approved the protocols and procedures.

Measurement of Health Characteristics and Oral Bisphosphonate Use

All women were personally interviewed regarding general health information at baseline. An inventory of all current, regularly used medications, including oral bisphosphonates, was taken at baseline and at 1, 3, and 6 years after random assignment for WHI-CT components. The same inventory was performed at baseline and 3 years after for WHI-OS. Participants were instructed to provide medication bottles or packaging for drugs taken at least twice per week during the previous 2 weeks. All medications were matched to the Medi-Span (Indianapolis, IN) Master Drug Data Base to ascertain detailed ingredient information. Intravenous bisphosphonates were not included.

Follow-Up for Endometrial Cancer Diagnoses

WHI participants reported diagnoses of invasive endometrial cancer, or hysterectomy for any reason, semi-annually (WHI-CT) or annually (WHI-OS). Incident endometrial cancers were adjudicated centrally by physicians through medical and pathology records review.¹⁵ Follow-up was censored at the earliest of the following events: time of hysterectomy, last known follow-up, or August 2010. The vast majority of endometrial cancers (81%) were diagnosed at local stage.

Study Exclusions

Women reporting a history of endometrial or breast cancer, hysterectomy, antiestrogen use, or hormone therapy specifically to treat bone fracture before study entry were excluded from analyses, as were women with missing information on endometrial cancer or hysterectomy at baseline or during follow-up. Requiring women to have an intact uterus at baseline led to the exclusion of all 10,739 women who participated in an estrogen-alone randomized trial. The final analytic study population included 89,918 WHI participants.

Statistical Analyses

We estimated adjusted hazard ratios (HRs) with 95% CIs from Cox regression models comparing bisphosphonate users to nonusers, treating bisphosphonate use as a time-varying never/ever variable, with baseline hazard stratified by WHI study component and adjustment for the following baseline covariates: age, 5-year hip fracture probability,¹⁸ body mass index (BMI), race, education, smoking status, estrogen-only use, estrogen-progestin use, oral contraceptive use, parity, and mammography. The 5-year hip fracture probability considers age, race, weight, height, self-reported health, diabetes, physical activity, bone fracture after age 54 years, parental hip fracture, smoking, and corticosteroid use. We considered further stratification by intervention arm of WHI-CT in sensitivity analyses. The proportional hazards assumption was assessed by testing for interactions with log-transformed time on study.

Bisphosphonate use was treated as a time-varying never/ever variable by updating baseline use at years 1, 3, and 6 for women in WHI-CT and at year 3 for women in WHI-OS. We permitted nonusers to become users over follow-up, but users could not become nonusers. We assessed interactions in the association between bisphosphonate use and endometrial cancer by age, BMI, and hip fracture probability score, each measured at baseline. All statistical tests were performed using SAS version 9.2 (SAS Institute, Cary, NC) and were two-sided, with P ≤ .05 considered statistically significant.

RESULTS

Of the 89,918 women included in our analyses, 39,261 (44%) were enrolled onto one or more of the three randomized trials, and 50,657 (56%) were exclusively in the WHI-OS. Bisphosphonate use at baseline was uncommon (2%) but had markedly increased by year 6 (10%). Alendronate was the most common type of bisphosphonate, accounting for more than 90% of use.

Users of any type of bisphosphonate at baseline were slightly older, more highly educated, less likely to be current smokers, and leaner than nonusers (Table 1). Users and nonusers were not that different in their use of hormone therapy before enrollment. Compared with nonusers, however, a lower percentage of bisphosphonate users at baseline participated in the combined estrogen plus progestin randomized trial (19% of nonusers and 11% of users, but treatment and placebo arms were well-balanced between users and nonusers).

Table 1.

Baseline Demographics and Clinical Characteristics of Women's Health Initiative Participants According to Use of Oral Bisphosphonates (1993 to 2010)

Characteristic	Oral Bisphosphonate User
	At Baseline				At Baseline, Year 1, or Year 3*
	No (n = 88,073)		Yes (n = 1,845)		No (n = 76,993)		Yes (n = 6,293)
	No.	%	No.	%	No.	%	No.	%
Age at eligibility screening, years
Mean	62.9		67.2		62.8		66.1
SD	7.2		6.4		7.1		6.6
50-59	31,084	35.3	229	12.4	27,627	35.9	1,091	17.3
60-69	39,009	44.3	887	48.1	34,119	44.3	3,144	50.0
≥ 70	17,980	20.4	729	39.5	15,247	19.8	2,058	32.7
Race/ethnicity
White	74,459	85.7	1,671	91.5	65,447	86.2	5,700	91.8
Black/African American	6,177	7.1	19	1.0	5,286	7.0	101	1.6
Hispanic/Latino	3,354	3.9	40	2.2	2,780	3.7	113	1.8
Asian/Pacific Islander	2,536	2.9	97	5.3	2160	2.8	290	4.7
American Indian/Alaskan native	319	0.4	0	0.0	269	0.4	6	0.1
Other/missing	1,228	—	18	—	1,051	—	83	—
Education
Less than high school	3,998	4.6	59	3.2	3,345	4.4	179	2.9
High school/vocational	22,129	25.3	425	23.2	19,428	25.4	1,433	23.0
Some college	23,097	26.4	398	21.7	20,157	26.4	1,475	23.6
College	10,561	12.1	278	15.2	9,305	12.2	868	13.9
Graduate/professional	27,656	31.6	671	36.6	24,233	31.7	2,285	36.6
Missing	632	—	14	—	525	—	53	—
Body mass index, kg/m2
Mean	27.6		25.1		27.7		25.1
SD	5.9		4.8		5.8		4.6
< 25	32,749	37.2	1,052	57.0	28,160	36.6	3,549	56.4
25 to < 30	29,939	34.0	562	30.5	26,343	34.2	1,939	30.8
≥ 30	24,576	27.9	222	12.0	21,794	28.3	753	12.0
Missing	809	—	9	—	696	—	52	—
Smoking
Never	43,833	50.4	990	54.3	38,564	50.7	3,309	53.3
Former	36,961	42.5	747	41.0	32,291	42.5	2,584	41.6
Current	6,145	7.1	85	4.7	5,195	6.8	316	5.1
Missing	1,134	—	23	—	943	—	84	—
Parity, No. of children
Nulliparous	11,131	12.7	284	15.5	9,639	12.6	906	14.5
1	7,718	8.8	148	8.1	6,658	8.7	512	8.2
2	22,040	25.2	472	25.7	19,245	25.2	1,625	25.9
3	21,026	24.0	434	23.7	18,464	24.1	1,539	24.6
≥ 4	25,598	29.3	497	27.1	22,514	29.4	1,681	26.8
Missing	560	—	10	—	473	—	30	—
Mammography in prior 2 years
No	14,929	17.5	152	8.5	12,966	17.4	617	10.1
Yes	70,410	82.5	1,644	91.5	61,739	82.6	5,503	89.9
Missing	2,734	—	49	—	2,288	—	173	—
Oral contraceptive use
Never	50,407	57.3	1,281	69.4	43,446	56.4	4,182	66.5
< 1 year	7,658	8.7	127	6.9	6,769	8.8	479	7.6
1 to < 3 years	7,865	8.9	129	7.0	6,972	9.1	452	7.2
≥ 3 years	22,115	25.1	308	16.7	19,784	25.7	1,180	18.8
Missing	28	—	0	—	22	—	0	—
E-alone use
Never	77,834	88.4	1,593	86.3	68,091	88.4	5,445	86.5
< 5 years	6,714	7.6	167	9.1	5,820	7.6	585	9.3
5 to < 10 years	1,763	2.0	44	2.4	1,551	2.0	129	2.0
≥ 10 years	1,761	2.0	41	2.2	1,530	2.0	134	2.1
Missing	1	—	0	—	1	—	0	—
E+P use
Never	54,057	61.4	1,162	63.0	46,709	60.7	4,001	63.6
< 5 years	17,097	19.4	353	19.1	15,154	19.7	1,134	18.0
5 to < 10 years	9,418	10.7	152	8.2	8,451	11.0	587	9.3
≥ 10 years	7,500	8.5	178	9.6	6,678	8.7	571	9.1
Missing	1	—	0	—	1	—	0	—
5-Year predicted probability of hip fracture, %†
< 0.5	63,449	72.0	757	41.0	56,157	72.9	3,213	51.1
0.5 to 1	13,334	15.1	481	26.1	11,511	15.0	1,459	23.2
≥ 1	11,290	12.8	607	32.9	9,325	12.1	1,621	25.8
Missing	0	—	0	—	0	—	0	—
Women's Health Initiative study components‡
Observational study	49,349	56.0	1,308	70.9	42,587	55.3	4,145	65.9
Calcium/vitamin D CT	20,844	23.7	226	12.2	19,211	25.0	1,008	16.0
Dietary modification CT	26,999	30.7	364	19.7	23,873	31.0	1,491	23.7
Hormone CT	16,329	18.5	200	10.8	14,670	19.1	818	13.0
Hormone CT arm
Not included in hormone CT	71,744	81.5	1,645	89.2	62,323	80.9	5,475	87.0
Placebo	7,960	9.0	105	5.7	7,076	9.2	479	7.6
E+P	8,369	9.5	95	5.1	7,594	9.9	339	5.4

Abbreviations: CT, clinical trial; E, estrogen; E+P, estrogen plus progestin; SD, standard deviation.

^*Among those remaining at risk of endometrial cancer at year 3.

^†Calculated as described in Robbins et al.¹⁸

^‡Percentages do not sum to 100% as a result of overlap of study components.

During study follow-up, 1,123 women were diagnosed with incident endometrial cancer (1,070 nonusers and 53 users of bisphosphonates). The crude incidence of endometrial cancer was 12 per 10,000 person-years of follow-up among nonusers and eight per 10,000 person-years among users of any type of bisphosphonate (seven per 10,000 person-years among alendronate users). Bisphosphonate use was inversely associated with age-adjusted endometrial cancer risk (HR, 0.76; 95% CI, 0.61 to 0.94; P = .01; Table 2). The inclusion of confounders, selected a priori, left the association largely unchanged (bisphosphonate users: HR, 0.80; 95% CI, 0.64 to 1.00; P = .05; alendronate users: HR, 0.77; 95% CI, 0.61 to 0.98; P = .03). In models without adjustment for variables already accounted for in the Robbin's fracture probability (age, race, smoking, and BMI), the HR for bisphosphonate use changed from 0.80 (95% CI, 0.64 to 1.00) to 0.74 (95% CI, 0.59 to 0.92).

Table 2.

Relative Risk of Endometrial Cancer Comparing Oral Bisphosphonate Users to Nonusers, Women's Health Initiative (1993 to 2010)

Oral Bisphosphonate Use*	Person-Years at Risk†	No. of Endometrial Cancer Events	Age-Adjusted Analysis			Multivariable-Adjusted Analysis
			HR‡	95% CI	P	HR‡§	95% CI	P
Any use
No	871,495	1,070	1.00		.01	1.00		.05
Yes	68,602	53	0.76	0.61 to 0.94		0.80	0.64 to 1.00
Type
Alendronate sodium	63,297	47	0.72	0.57 to 0.90	.005	0.77	0.61 to 0.98	.03
Risedronate sodium	3,349	5	1.33	0.73 to 2.42	.34	1.44	0.77 to 2.70	.25
Other‖	1,956	1	0.37	0.05 to 2.59	.31	0.40	0.06 to 2.82	.36
Duration of use, years
< 1	22,101	17	0.73	0.50 to 1.08	.11	0.85	0.57 to 1.25	.40
1 to 3	27,817	19	0.75	0.54 to 1.05	.09	0.81	0.58 to 1.15	.24
≥ 3	18,684	17	0.77	0.53 to 1.11	.16	0.76	0.51 to 1.13	.17

Abbreviations: HR, hazard ratio; WHI-CT, Women's Health Initiative clinical trials; WHI-OS, Women's Health Initiative observational study.

^*Users reported at least 2 weeks of use; nonusers include never-users and those who used for < 2 weeks. Baseline oral bisphosphonate use was updated at years 1, 3, and 6 for women in WHI-CT and at year 3 for women in WHI-OS.

^†Median follow-up time of 12.5 years.

^‡Baseline hazard stratified by Women's Health Initiative study component.

^§Adjusted for age, 5-year hip fracture probability, body mass index, race, education, smoking status, estrogen-only use, estrogen-progestin use, oral contraceptive use, parity, and mammography, all measured at baseline.

^‖Other includes etidronate disodium and tiludronate disodium.

The magnitude of HRs was similar for different durations of use, but the HRs were less precise than the estimate ignoring duration of use. Results were largely unchanged in models that also stratified the baseline hazard by intervention arm of each randomized trial. We found no statistically significant interactions with BMI (P = .41), age at baseline (P = .21), or hip fracture probability score (P = .83).

DISCUSSION

In this large cohort of postmenopausal women, oral bisphosphonate use was associated with a statistically significant reduction in endometrial cancer risk. The mechanisms underlying the observed relationship are unclear. Bisphosphonates prevent metastasis to bone,¹⁹ and numerous in vivo studies have demonstrated their proapoptotic and antiangiogenic properties.^20,21 This association might be hormonally mediated, especially given that bisphosphonate history is most strongly, if not exclusively, associated with lower risk of hormone receptor–positive breast cancer as shown in the study of contralateral breast cancer by Monsees et al.⁶

To our knowledge, this is the first prospective study to evaluate bisphosphonate use and endometrial cancer risk. The relationship has been evaluated in the United Kingdom General Practice Database, where the findings were inconsistent,^11,12 and the Cancer in the Ovary and Uterus Study (CITOUS), which found that the use of bisphosphonates was associated with reduced risk of endometrial cancer.¹³ The 20% reduction in risk we observed is far more modest than the 60% reduction reported in CITOUS. Although CITOUS determined bisphosphonate use from pharmacy records, limiting the potential for recall bias, this bias could have been introduced from the measurement of other covariates, which were collected retrospectively from self-report.

The results presented here have limitations. This analytic study design is observational rather than randomized.²² Therefore, confounding is a possibility because women using oral bisphosphonates may have done so because of their high fracture risk as a result of low endogenous estrogen, perhaps as a consequence of low weight, which could place them at low endometrial cancer risk.²³ We carefully controlled for weight and considered various approaches to define this risk factor; all seemed to provide similar control for any residual confounding with little change in our measures of association. We also took fracture risk, as summarized by the index of Robbins et al,¹⁸ into account. This fracture probability has proved to be a valid measure of fracture risk, although it does not include measured bone mineral density. Thus, our full-adjustment approach to protect against residual confounding leads us to be conservative about the strength of the association. Finally, although this is a large study, our numbers were limited to explore heterogeneity in associations according to endometrial cancer subtypes. Given that type II endometrial tumors are less likely to be hormonally mediated than type I tumors,²⁴ it is plausible that bisphosphonate use would be most relevant to risk of type I tumors; however, the modest number of type II tumors in the study population (approximately 10%) precluded separate evaluation by tumor type.

The prevalence of bisphosphonate use was initially low in this cohort. However, we were able to incorporate the substantial increase in use with a time-varying variable. All study participants were instructed to present regularly used medications during the scheduled drug inventories. Any misclassification or under-reporting of exposure history is unlikely to differ between women who developed endometrial cancer during follow-up and those who did not; nondifferential misclassification would have yielded attenuated measures of the true association.²⁵

The strengths of our large study include the prospective collection of detailed data on endometrial cancer risk factors, accounting for prior and incident hysterectomy, and centralized adjudication of endometrial cancer diagnoses. In summary, our findings suggest that use of bisphosphonates is modestly associated with reduced endometrial cancer risk, a finding consistent with the inverse association between use of this medication and breast cancer risk.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Disclosures provided by the authors are available with this article at www.jco.org.

AUTHOR CONTRIBUTIONS

Conception and design: Polly A. Newcomb

Financial support: Polly A. Newcomb

Administrative support: Polly A. Newcomb, Jean Wactawski-Wende

Provision of study materials or patients: Jean Wactawski-Wende

Collection and assembly of data: Polly A. Newcomb, Garnet L. Anderson, Jean Wactawski-Wende, Rowan T. Chlebowski

Data analysis and interpretation: Polly A. Newcomb, Michael N. Passarelli, Amanda I. Phipps, Garnet L. Anderson, Gloria Y.F. Ho, Mary Jo O'Sullivan, Rowan T. Chlebowski

Manuscript writing: All authors

Final approval of manuscript: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Oral Bisphosphonate Use and Risk of Postmenopausal Endometrial Cancer

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or jco.ascopubs.org/site/ifc.

Polly A. Newcomb

No relationship to disclose

Michael N. Passarelli

No relationship to disclose

Amanda I. Phipps

No relationship to disclose

Garnet L. Anderson

No relationship to disclose

Jean Wactawski-Wende

No relationship to disclose

Gloria Y.F. Ho

No relationship to disclose

Mary Jo O'Sullivan

No relationship to disclose

Rowan T. Chlebowski

Consulting or Advisory Role: Novartis, Pfizer, Genentech, Novo Nordisk, Genomic Health

Speakers' Bureau: Novartis

Travel, Accommodations, Expenses: Novartis, Pfizer, Genentech, Novo Nordisk, Genomic Health