Bisphosphonates are widely prescribed to increase bone density in postmenopausal women with osteopenia or osteoporosis. Aminobisphosphonates have numerous anticancer properties and reduce bone metastases in cancer patients. Several studies, including the Women's Health Initiative (WHI), have found that use of oral bisphosphonates is associated with reduced risk of developing breast cancer, but less is known about associations with other common malignancies in women such as colorectal cancer (CRC). A few case-control and retrospective cohort studies have reported decreased risk of CRC among bisphosphonate users. In contrast, a prospective cohort study found no association. We evaluated the association between oral bisphosphonate use and CRC incidence in 156,826 postmenopausal women, ages 50 to 79 years, who participated in WHI clinical trials and observational study. 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. A total of 1931 women were diagnosed with incident invasive CRC during a median follow-up of 12 years. Alendronate was the most commonly used bisphosphonate, accounting for >90% of the total person-years of use. The association between oral bisphosphonate use and CRC risk did not reach statistical significance (hazard ratio [HR] from multivariable-adjusted models = 0.88; 95% confidence interval [CI] 0.72–1.07; p = 0.19). Furthermore, we did not observe greater risk reductions for women with longer duration of use. Uncontrolled confounding may explain why previous studies have observed an association.
Bisphosphonates are widely used to treat bone loss, with more than 150 million prescriptions dispensed in the United States from 2005 to 2009. Aminobisphosphonates prevent the attachment of osteoclasts to bone by inhibiting farnesyl-pyrophosphate synthase, and experimental studies have demonstrated their ability to promote apoptosis and suppress growth and angiogenic factors. Bisphosphonates prevent bone metastases, but it is unclear if they play a role in the primary prevention of cancer. Use of bisphosphonates has been found to be associated with ∼30% reduced risk of breast cancer in several studies,[5, 6] including the Women's Health Initiative (WHI). Studies of bisphosphonates and the risk of developing other hormone-related cancers such as colorectal cancer (CRC) have been inconsistent.
The Molecular Epidemiology of Colorectal Cancer (MECC) case-control study reported nearly 60% reduced risk of CRC for postmenopausal women who used bisphosphonates for ≥1 year (odds ratio [OR] = 0.41; 95% confidence interval [CI] 0.25–0.67). More modest inverse associations were observed in two studies nested within the United Kingdom General Practice Research Database (GPRD)[9, 10] (OR = 0.87; CI 0.77–1.00 and OR = 0.74; CI 0.60–0.91), and one from the Manitoba Drug Program Information Network (OR = 0.78; CI 0.65–0.94). Analyses using the Danish National Prescription Database are conflicting; one reported an inverse association (hazard ratio [HR] = 0.69; CI 0.60–0.79), but another suggested no association (HR = 1.16; CI 0.93–1.45). To date, the only data from a prospective cohort study come from the Nurses' Health Study (NHS), which observed no association (HR = 1.04; CI 0.82–1.33). Given the varied findings of previous studies, we investigated the association between oral bisphosphonate use and CRC incidence in participants of the WHI clinical trials (WHI-CT) and observational study (WHI-OS).
Materials and Methods
WHI enrolled 161,808 postmenopausal women, 50 to 79 years of age, from 40 clinical centers across the United States from 1993 to 1998. WHI-OS participants were generally similar to those of WHI-CT but were ineligible or unwilling to be included in a randomized trial. Study eligibility and implementation details have been extensively documented.
Measurement of health characteristics and oral bisphosphonate use
All women were interviewed on general health information at baseline. Total calcium and vitamin D intake were estimated from two sources: 1) diet using a semiquantitative food-frequency questionnaire, and 2) from an inventory of medications and supplements. An inventory of all current, regularly used medications, including bisphosphonates, was taken at baseline and at 1, 3, and 6 years after randomization for WHI-CT components. The same inventory was performed at baseline and 3 years after for WHI-OS. Participants were instructed to bring medication bottles/packaging for drugs taken at least twice per week during the previous 2 weeks. All medications were matched to the Medi-Span (Indianapolis, IN, USA) Master Drug Data Base (MDDB) to ascertain detailed ingredient information.
Follow-up for colorectal cancer diagnoses
Clinical outcomes, including diagnoses of CRC, were reported by participants every 6 months for WHI-CT and annually for WHI-OS. Study physicians adjudicated case reports by reviewing medical records and pathology reports. Participants reported receipt of colorectal exams for any indication, including hemoccult test, sigmoidoscopy, and colonoscopy. Diagnoses of colorectal polyps were also documented, but these were not physician-adjudicated.
The estrogen-progestin trial was stopped in 2002, the estrogen-alone trial was stopped in 2004, and the primary study period for all other components ended in 2005, after which time 115,400 participants reconsented to be followed for additional outcomes including diagnoses of CRC.
We excluded 5 women who had not turned 50 years old at baseline, 946 with a personal history of colon, rectal, bowel, or intestinal cancer, 1760 with a personal history of ulcerative colitis, and 2271 who reported ever having part of their intestine removed. Among the eligible 156,826 women, a total of 2070 reported a diagnosis of CRC during follow-up through August 2009. We did not count 139 diagnoses as outcomes (70 that could not be adjudicated and 69 that were adenocarcinoma in situ), leaving 1931 with CRC.
We calculated descriptive statistics at baseline and also for any use during the first 3 years of the study. HRs and CIs comparing bisphosphonate users to nonusers were estimated using proportional hazards regression. 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 WHI-OS (nonusers could become users over time, but users could not become nonusers). All models stratified the baseline hazard by study component and randomization arm for the estrogen-alone trial, estrogen-progestin trial, and calcium/vitamin D trial. The proportional hazards assumption was assessed by testing for interactions with log-transformed time-on-study.
Regression models included adjustment for covariates measured at baseline (not time-varying) as categorized in Table 1: age, race, education, body mass index (BMI), smoking status, alcohol consumption, physical activity, aspirin and other nonsteroidal anti-inflammatory drug (NSAID) use, first-degree family history of CRC, estrogen-only use, estrogen-progestin use, mammography history, lower endoscopy (sigmoidoscopy or colonoscopy) history, total calcium intake, total vitamin D intake, and 5-year fracture probability as predicted by the model developed by Robbins and colleagues. This fracture probability is based on 11 factors (age, race, weight, height, self-reported health, diabetes, physical activity, bone fracture after age 54, parental hip fracture, smoking, and corticosteroid use) and has been shown to correlate well with bone mineral density (BMD). We also explored associations with bisphosphonate type and duration of use as time-varying variables. Analyses were performed using SAS 9.2 (SAS Institute, Inc., Cary, NC, USA). The p values are two-sided with p ≤ 0.05 considered statistically significant.
|Characteristic at baseline||Oral bisphosphonate user at baseline||Oral bisphosphonate user at baseline, year 1, or year 3a|
|No (n = 153,782)||Yes (n = 3044)||No (n = 133,304)||Yes (n = 10,031)|
|WHI-CT: calcium/vitamin D||35,134||23||347||11||32,317||24||1514||15|
|WHI-CT: dietary modification||47,172||31||573||19||41,590||31||2289||23|
|American Indian/Alaska native/other/unknown||2833||2||34||1||2385||2||134||1|
|Some college/college graduate||59,070||39||1116||37||51,299||39||3773||38|
|Body mass index (kg/m2)|
|25 to <30||53,026||35||967||32||46,334||35||3154||32|
|Current <1 drink/week||50,352||33||930||31||43,782||33||3115||31|
|Current 1 to <7 drinks/week||39,243||26||811||27||34,238||26||2853||29|
|Current ≥7 drinks/week||17,794||12||381||13||15,646||12||1323||13|
|Recreational physical activity (MET-hours/week)|
|0 to <5||54,114||37||857||28||46,493||37||2913||30|
|5 to <10||25,405||17||512||17||21,995||17||1711||17|
|10 to <30||52,075||36||1290||43||45,334||36||4099||42|
|Regular NSAID use||29,309||19||608||20||25,660||19||1785||18|
|Regular aspirin use||32,941||21||789||26||28,674||22||2449||24|
|Family history of CRC in first-degree relatives||23,061||16||488||18||19,963||16||1583||17|
|5 to <10 years||10,855||7||165||5||9582||7||556||6|
|5 to <10 years||10,953||7||200||7||9740||7||710||7|
|Mammogram in prior 2 years||124,109||83||2669||91||108,105||84||8749||90|
|Last colorectal endoscopy|
|Never had endoscopy||72,325||50||1101||37||62,458||50||3877||40|
|<5 years ago||45,866||32||1330||44||39,796||32||4038||41|
|≥5 years ago||27,493||19||578||19||23,830||19||1909||19|
|Last hemoccult test|
|Never had hemoccult test||37,054||26||471||16||31,667||25||1676||17|
|<5 years ago||80,861||56||2031||68||70,399||56||6528||67|
|≥5 years ago||27,585||19||490||16||23,853||19||1609||16|
|Ever had colon polyps removed||11,993||8||368||12||10,222||8||1051||11|
|Total calcium intakec (mg/day)|
|800 to <1,200||37,423||24||583||19||32,691||25||2169||22|
|Total vitamin D intakec (IU/day)|
|400 to <600||37,793||25||824||27||32,771||25||2759||28|
|History of osteoporosis||9443||6||2166||72||8525||6||2024||71|
|5-year predicted probability of hip fractured (%)|
|0.5 to 1||23,835||16||767||25||20,450||15||2305||23|
Bisphosphonate use was low at baseline (n = 3044 users, 2%) but became increasingly common over the course of follow-up (10,349 users at year 3, 7% of those remaining at risk of CRC, of whom 3373 used <1 year). Consistent with the indication for use, 72% of baseline users reported a physician diagnosis of osteoporosis compared with only 6% of nonusers (Table 1). Baseline users were more likely to be older, white, of lower BMI, nonsmokers, and to have recently undergone a cancer-screening test. Users also had higher total calcium intake, vitamin D intake, and 5-year probability of hip fracture. Women who reported use anytime during the first 3 years of the study (including baseline) were generally similar to baseline users but, on average, were slightly younger, had lower calcium, vitamin D, and 5-year probability of hip fracture—trends that likely reflect the prescription of bisphosphonates to a broader patient population. Over the course of the study, use of the aminobisphosphonate alendronate accounted for >90% of the total person-time of bisphosphonate use.
Based on a multivariable-adjusted model, the estimated HR of CRC risk comparing bisphosphonate users to nonusers was not different from the null at the α = 0.05 level of significance (HR = 0.88; CI 0.72–1.07; p = 0.19; Table 2). This fully adjusted estimate was attenuated from an age-only adjusted estimate. Because alendronate use predominated, type-specific estimates for risedronate and other types (etidronate and tiludronate) were not very precise. The magnitude of the association was stronger for recent adopters of bisphosphonates than for longer-term users. We explored whether the association between CRC risk and bisphosphonate use depended on BMI, family history of CRC, hormone use by type, vitamin D and calcium intake, cancer-screening tests, and indication for bisphosphonate use. The strongest associations were among women with low vitamin D intake (<400 IU/day; HR = 0.67; CI 0.49–0.92) and who had family history of CRC (HR = 0.59; CI 0.34–1.00). Neither interaction, however, achieved statistical significance with consideration of multiple tests (p interaction = 0.05 and 0.16, respectively). Replacing the 5-year predicted probability of hip fracture with all of its components as actual adjustment variables also did not substantially alter estimates.
|Oral bisphosphonate usea||Person-years at risk||No. of CRC events||Age-adjusted||Multivariable-adjusted|
|HRb (95% CI)||p||HRb,c (95% CI)||p|
|No||1,555,268||1805||1.00 (Reference)||1.00 (Reference)|
|Yes||111,538||126||0.81 (0.67, 0.97)||0.02||0.88 (0.72, 1.07)||0.19|
|Alendronate sodium||102,219||115||0.78 (0.64, 0.95)||0.01||0.84 (0.68, 1.04)||0.11|
|Risedronate sodium||5666||7||0.91 (0.43, 1.91)||0.80||0.93 (0.42, 2.09)||0.86|
|Otherd||3653||4||0.78 (0.29, 2.08)||0.62||0.72 (0.23, 2.23)||0.57|
|Duration of use|
|<1 year||36,692||29||0.64 (0.45, 0.91)||0.01||0.70 (0.49, 1.02)||0.06|
|1 to 3 years||44,206||59||0.89 (0.68, 1.18)||0.42||0.89 (0.65, 1.20)||0.43|
|≥3 years||30,640||38||0.79 (0.56, 1.12)||0.19||0.95 (0.67, 1.35)||0.78|
We did not find sufficient evidence to conclude that there was a meaningful association between bisphosphonate use and CRC risk in WHI. If any association exists, it may be smaller than what our study could detect. If bisphosphonates reduced CRC risk, we would expect the relation to be stronger for women using for longer durations. This was not reflected in our data. Instead, any apparent risk reduction may be the consequence of recent physician contact.
Our results are consistent with those of Khalili and colleagues, where no association was observed over 10 years of follow-up in NHS (HR = 1.04; CI 0.82–1.33). There are, however, some differences between their analysis and ours. WHI was twice as large (1931 CRC cases among 156,826 women compared with 801 CRC cases among 86,277 women at risk in NHS) and confirmed bisphosphonate use, type, and nonintravenous preparation via assessments of medication bottles and not simply from self-reported questionnaire. Among other variables, we adjusted HRs for a predicted fracture probability (BMD surrogate), recent physician contact at mammography, education, and hormone type, which were not part of the NHS analysis. Khalili and colleagues, however, adjusted for some exposures that we did not, including regular statin use, folate intake, red meat intake, and personal history of colorectal polyps beyond receipt of endoscopy.
Several previous case-control and retrospective studies have observed a reduced risk of developing CRC among users.[8-12] Analyses from pharmacy databases generally have limited ability to adequately control for confounding. The Danish studies did not adjust for BMI, calcium, or vitamin D. One GPRD study adjusted for calcium and vitamin D but included only prescription sources. None, including MECC, measured calcium and vitamin D intake from diet.
Our study is not without limitations. WHI began around the time bisphosphonates first started being commonly prescribed. Use was low at baseline (2%), and half of these women were users for <1 year. Although the prevalence of use increased during follow-up, of the women who developed CRC, fewer than 150 had ever used bisphosphonates during the first 6 years of the study. Given that nearly all users were taking alendronate, we had little power to detect associations for other types. Consequently, we could not confirm the previously reported inverse association specific to risedronate. It is possible that any benefits of bisphosphonates for CRC prevention require higher doses or longer durations than that used by women in WHI. We did not evaluate dose and could not account for continued compliance between medication assessments. Further studies may provide a more complete evaluation.
Despite the promising experimental evidence of the antiproliferative action of aminobisphosphonates, our study lends further support to the conclusion that use of bisphosphonates does not meaningfully alter the risk of developing CRC. Women who are prescribed bisphosphonates may have bone loss that results from factors that are also associated with CRC risk, including lifetime hormonal exposures. Any antiproliferative action of bisphosphonates may be offset by the increased risk of CRC in these women. Our results suggest that uncontrolled confounding may explain why previous studies have observed decreased risk of CRC among bisphosphonate users.
MP, PN, AL, DL, and GH state that they have no conflicts of interest. RC has served as a paid consultant for AstraZeneca and has received honoraria from AstraZeneca, Novartis, and Amgen. All authors state that they have no restrictions on access to raw data or statistical analyses.
The authors thank the WHI investigators and staff for their skill and dedication, and the study participants for making the program possible. A listing of WHI investigators can be found at https://cleo.whi.org/researchers/Documents%20%20Write%20a%20Paper/WHI%20Investigator%20Short%20List.pdf.
The WHI program is funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services, through contracts HHSN268201100046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, HHSN268201100004C, and HHSN271201100004C. MP is supported by a training grant from the National Cancer Institute (T32-CA009168).
Authors' roles: Study design: AL, DL, GH, and RC. Study conduct: AL, DL, GH, and RC. Data collection: AL, DL, GH, and RC. Data analysis: MP and PN. Data interpretation: All. Drafting manuscript: MP and PN. Revising manuscript content: All. Approving final version of manuscript: All. MP takes responsibility for the integrity of the data analysis.