Combination Therapy With Hormone Replacement and Alendronate for Prevention of Bone Loss in Elderly Women:  A Randomized Controlled Trial FREE

The impact of osteoporosis is most pronounced in elderly women who have the greatest risk of fracture.^1- 2 Available antiresorptive agents increase bone mineral density (BMD) and reduce fractures in women with postmenopausal osteoporosis.³ Although the relationship between BMD and fracture reduction is not linear,^4- 6 meta-analyses of multiple clinical trials of antiresorptive therapies reveal that an increase in BMD correlates well with the reduction in the rate of vertebral and nonvertebral fractures.^6- 8 Furthermore, Hochberg et al⁷ demonstrated that women taking the bisphosphonate alendronate who had the greatest improvement in vertebral BMD also had the greatest reduction in vertebral fractures.

The rationale for combining 2 antiresorptive agents with different mechanisms of action is to induce greater increases in bone mass and, hopefully, greater reduction in fractures than would be observed with a single antiresorptive agent. A 2-year study by Bone et al⁹ demonstrated that BMD was greater in younger women taking combination therapy than it was in women taking either estrogen or alendronate alone. Fracture reduction was not a primary end point, and no study of combination antiresorptive therapy has had the power to detect a reduction in risk of fracture corresponding to higher bone mass. Moreover, it is not known if improvements in bone mass with combination therapy are relevant to less healthy, older women treated with a standard replacement regimen of estrogen and medroxyprogesterone over 3 years. It is also not known if combination therapy is safe or well tolerated in this group of older women, or if compliance will differ between groups. Finally, it is unknown whether results would be the same with hormone replacement plus bisphosphonates. Therefore, our study was designed to examine the efficacy and safety of combination therapy with hormone replacement and alendronate, compared with each agent alone in a group of community-dwelling, elderly women.

We screened 573 community-dwelling women aged 65 or older from the greater Boston area for a single-center clinical trial. Participants were excluded if they had a history of illnesses that could affect bone mineral metabolism (eg, current hyperthyroidism or hyperparathyroidism, renal failure, hepatic failure, and active malignancy) or if they were currently taking medications known to alter bone mineral metabolism (eg, glucocorticoids, anticonvulsants, excess thyroid hormone). Participants were also excluded if they had been treated with osteoporosis medications (eg, bisphosphonates, hormone replacement, or calcitonin) within a year of screening. In addition, women were excluded if they had any contraindications for hormone replacement or alendronate or had a baseline femoral neck BMD of 0.9 g/cm² or greater (ie, zero SD of mean peak BMD using Hologic database prior to the Third National Health and Nutrition Examination Survey database¹⁰). The protocol was approved by the institutional review board at the Beth Israel Deaconess Medical Center in Boston, Mass. Participants were advised of the nature of the study and provided written informed consent prior to participation.

This study used a randomized, double-blind, 2 × 2 factorial design and compared the effect of alendronate or placebo combined with hormone replacement or placebo. We conducted the study from January 1996 through May 2001. Women received alendronate sodium, 10 mg/d (Merck Research Laboratories, Rahway, NJ), or matching placebo. For hormone replacement, women who had had a hysterectomy were given 0.625 mg/d of conjugated equine estrogen (Premarin, Wyeth Ayerst, Philadelphia, Pa) or matching placebo, and women with an intact uterus received conjugated equine estrogen, 0.625 mg/d, with medroxyprogesterone, 2.5 mg/d, (PremPro, Wyeth Ayerst) or matching placebo. A standard food frequency questionnaire was administered, and participants with a daily calcium intake of less than 1000 mg from dietary and supplementary sources were provided with 600 mg per tablet of elemental calcium and 200 IU per tablet of vitamin D (Caltrate Plus D, Wyeth Ayerst) to increase their total daily intake to more than 1000 mg/d.¹¹ In addition, all participants received a daily multivitamin (400 IU per tablet) so that their daily vitamin D intake was 400 to 800 IU.

Of the 573 women screened, 485 completed screening (Figure 1). To minimize discontinuation following randomization, participants entered a 3-month, open-label, run-in phase prior to randomization, involving hormone replacement, alendronate placebo, calcium (if necessary), and a multivitamin. The rationale behind the run-in phase was to ensure that the 4 groups would be roughly equal after 3 years, such that women taking hormone replacement could be compared with women taking alendronate, combination therapy, or placebo. Three hundred seventy-three women, aged 65 to 90 years, completed the run-in phase and agreed to continue in the study (Figure 1). Reasons for discontinuation during the run-in phase were previously reported.¹² To summarize, 73 participants discontinued secondary to adverse events related to hormone replacement and 39 discontinued for reasons unrelated to hormone replacement. Participants completing the run-in phase and agreeing to continue in the study were then randomized to 1 of 4 treatment arms: (1) placebo only (hormone replacement placebo and alendronate placebo), (2) hormone replacement only (hormone replacement and alendronate placebo), (3) alendronate only (hormone replacement placebo and alendronate), or (4) combination therapy (hormone replacement and alendronate). All participants continued taking calcium and vitamin D supplementation throughout the study.

After the participant agreed to continue, the study coordinator entered the participant's data into a computer program and received the participant's randomization number. Using randomization lists prepared by the study statistician, the research pharmacy determined treatment using the participant's randomization number. Use of this computer program ensured that participants were randomized only after all procedures were completed and that participants were randomized in strict sequential order at the time they had completed all procedures. Randomization was stratified by prior history of hysterectomy and 3 levels of total hip BMD to ensure balance between groups. Within each strata, randomization was blocked using block sizes of 4, 8, or 12 (block size randomly determined) to reduce the chance that study staff could deduce the treatment assignment. Only the research pharmacist and study statistician had access to the complete randomization code. Sealed, light-proof envelopes were prepared for each participant with their hormone replacement assignment for use by the clinic gynecologist who was not involved in the study outcome activities. Four envelopes were reviewed by the clinic gynecologist during the protocol for safety or symptom management. The clinic gynecologist was instructed not to share this information either with the participant or with study staff. One additional envelope was opened by a study staff member in error. All other envelopes were returned to the statistical staff at the end of the study with no evidence of tampering. The study statistician provided complete treatment information to nonstudy physicians during the course of the study for 3 participants after they had withdrawn from the study; this information was not shared with any study staff. Considerable effort was made to maintain blinding of participants. However, we did not attempt to assess the adequacy of blinding during the study.

Bone Mineral Density . Bone mineral density of the hip (total hip, femoral neck, trochanter, intertrochanter, and Ward triangle), lumbar spine (posteroanterior and lateral), and radius (ultradistal, mid-third, and one-third distal radius) were measured by dual-energy X-ray absorptiometry ([DXA] QDR 4500A densitometer; Hologic, Inc, Bedford, Mass) at baseline, randomization, and 6-month intervals for 3 years. As previously reported, the coefficients of variation of BMD in elderly women (mean [SD] age, 71 [7] years) using our densitometer were 1.7% for lateral lumbar spine, 1.5% for posteroanterior lumbar spine, 1.2% for total hip, and 1.9% for femoral neck.^13- 14 Quality control for bone density scans was performed by Synarc, Inc (San Francisco, Calif), which included visual inspection of the scans for correct scanning and analysis and a review of the quality control database using multirule Shewart charts.¹⁵

Clinical Characteristics. Weight was measured at baseline and every 6 months (Acme Digital In-Bed Scale, model 0501; Acme Medical Scale Co, San Leandro, Calif). Height was measured to the nearest millimeter 3 times per visit and the mean was calculated (Harpenden stadiometer; Holtrain Ltd, Crymych, Dyfed, United Kingdom). Body mass index (BMI) was calculated as kilograms per square meter. We assessed daily activity with the 27-point Instrumental Activities of Daily Living from Laughton and Brody,¹⁶ which measures a set of behaviors, including telephoning, shopping, food preparation, housekeeping, laundering, use of transportation, use of medicines, and financial behavior. We also administered the 30-point Folstein Mini-Mental State examination,¹⁷ which assesses cognitive function.

A priori, we specified that the study needed to randomize 92 participants to each treatment group (total 368 participants) to provide 80% power to detect a difference of 3% (or 90% power to detect a difference of 4% between each treatment group and the placebo group, with α = .05 [2-sided]). These calculations were based on variance estimates found in a previous study at our site¹⁸ and incorporated a 24% drop-out rate (8% per year) during the 3 years after randomization.

Analyses used treatment as assigned (ie, the intention-to-treat framework). Results are presented as mean (SD) in text and tables. Figures display mean (SEM). For comparisons of baseline values among all 4 groups, we used the Kruskal-Wallis test or χ² test as appropriate. Percentage changes from baseline at initiation of the run-in phase were calculated from participants with the measurement, without imputation for missing values. Statistical significance for changes from baseline within a group was tested using the Wilcoxon rank sum test. For comparisons between groups, we used a Bonferroni correction to adjust for the 3 primary treatment comparisons in the analysis: combined therapy vs alendronate alone, combined therapy vs hormone replacement alone, and alendronate alone vs hormone replacement alone. Binary variables (eg, clinical response) were compared using the Fisher exact test. For continuous variables (eg, BMD measures), we present results from a mixed-models analysis of variance of percentage change from baseline. We did not impute missing data because the mixed-models analysis of variance does not require complete data on each participant. Our analysis adjusted for time trend using date of bone density scan and an AR(1) correlation structure between measurements within a participant. The rate of change per year in BMD was calculated from this mixed-models analysis of variance separately for each group. For comparisons between 2 treatment groups, our initial model adjusted for time trend, included an indicator variable to allow for differences between the intercepts (value at the end of the run-in phase) for the 2 treatment groups, and included a time × treatment group interaction. Conceptually, this model is equivalent to fitting 2 nonparallel lines with different intercepts in an analysis of covariance. If the indicator variable for differences in the intercept was not statistically significant (P>.05), it was excluded and differences in time trend were assessed from a final model, including an overall time trend and the time × treatment group interaction. Similar results (not shown) were found when we imputed missing data using a last-value-carried-forward approach. The response to therapy was analyzed from participants who had a measurement at 36 months. A participant was considered a responder if, after 3 years, the change in BMD at the spine or hip was greater (more positive) than −1.0%.¹⁹ Cochran Mantel-Haenszel statistics were used to test for a trend over age groups adjusted for treatment groups. All analyses were performed using SAS version 8.1 (SAS Institute, Cary, NC).

No statistically significant differences were found in baseline characteristics or BMD between the 4 groups (Table 1). Baseline levels of serum calcium, albumin, parathyroid hormone, and 25-hydroxyvitamin D were within the normal range (Table 1). The mean (SD) dietary intake of calcium was 890 (437) mg/d and mean (SD) vitamin D intake was 252 (230) IU/d. Thirty-two percent of participants (120/373) were taking nonsteroidal anti-inflammatory medications or aspirin. Thirty-five percent (130/373) of women had had a hysterectomy and the distribution between estrogen or estrogen plus progesterone (or placebo) was similar across groups. The mean (SD) bone mass of the entire cohort was in the osteopenic classification by World Health Organization criteria; 34% (128/373) of the group had osteoporosis.^20- 21 Compliance and retention were similar in all groups. Fifty-one percent (48/94) to 63% (59/93) of participants were adherent, which was defined as taking 80% of both medications during the study (Table 2) (criteria previously cited in other estrogen trials^22- 23). Retention was 90% (337/373) for all participants after 3 years (Figure 1).

After 3 years,total hip BMD showed a mean (SD) increase of 5.9% (3.8) with combination therapy, an increase of 4.2% (3.8) with alendronate, and an increase of 3.0% (4.9) with hormone replacement (all P<.001), with maintenance of BMD in the placebo group. Similar trends were observed for the BMD of the femoral neck and trochanter (Figure 2). After 12 months and for the remainder of the study, the percentage change in total hip BMD was significantly greater in each of the 3 active treatment groups than in the placebo group (P<.001). At 36 months, the women treated with combination therapy had a significantly greater increase in total hip BMD than those on either alendronate or hormone replacement alone (P<.01). The rate of change of BMD (percentage per year) at the total hip was greater with combination therapy than with either active therapy alone (P<.01) (Table 3). These differences were statistically significant for combined therapy vs hormone replacement alone at all hip sites and for combined therapy vs alendronate alone for the total hip and trochanter.

At 3 years, women in the treatment groups had significant increases in posteroanterior and lateral lumbar spine BMD (P<.001), with maintenance of vertebral BMD in the placebo group (Figure 2). At all time points after randomization, BMD was greater in each treatment group than in the placebo group. At the posteroanterior lumbar spine, mean (SD) change BMD increased 10.4% (5.4) in the combination group, 7.7% (5.2) in the alendronate group, 7.1% (5.8) in the hormone replacement group (all P<.001), and 1.1% (4.6) in the placebo group (P<.05). Mean (SD) changes at the lateral lumbar spine reflected changes at the posteroanterior lumbar spine, with an increase of 11.8% (6.8) in women receiving combination therapy. The gain in vertebral BMD and the rate of change was greater in women taking combination therapy than either agent alone.

At 3 years, we observed small increases in mean (SD) BMD (1.9% [7.5]) at the ultradistal radius in women in the combination therapy group, but the changes in mean (SD) BMD remained stable at this site in the alendronate only and hormone replacement only groups, and decreased in the placebo group (−3.0% [5.8]; Figure 2). The BMD for the one-third distal radius remained stable in the active treatment groups but decreased in the placebo group (data not shown).

When the analysis comparing alendronate with hormone replacement therapy was performed as a head-to-head analysis without adjustment for multiple comparisons, BMD of the total hip, trochanter, intertrochanter, and Ward triangle were significantly higher following therapy with alendronate (P<.05) compared with hormone replacement therapy. We found no difference in BMD for femoral neck, vertebral, or wrist BMD between the 2 active groups. The effects of combined hormone replacement and alendronate were not additive but were less than the sum of the effect of each agent alone (P<.001 for total hip, trochanter, intertrochanter, and lateral lumbar spine; P<.05 for femoral neck and posteroanterior lumbar spine). The changes in BMD were not consistently related to baseline BMD, age, BMI, dietary calcium or vitamin D, or baseline serum vitamin D. The changes in BMD were related to adherence; BMD was higher in participants who were adherent in all 3 treatment groups (data not shown).

Most women receiving active treatment responded to therapy, but significant differences were found in the response rate between the 3 active groups (Table 2). The combination therapy group had a significantly higher response rate than hormone replacement alone for the total hip (P<.05) but not for the lumbar spine. The women taking alendronate alone had a higher response rate at the intertrochanter (P<.05) than women taking hormone replacement alone. No significant differences were found between BMD increases for women taking estrogen plus progesterone vs women taking unopposed estrogen (data not shown). The response to therapy was unrelated to baseline BMD, BMI, hysterectomy, dietary calcium and vitamin D intake, baseline vitamin D levels, or use of NSAIDs or aspirin. However, at the total hip, trochanter, and intertrochanter, the response was related to age at consent, with older participants responding less often. For example, in the 3 active treatment groups, the proportion responding at the total hip was 93% in women younger than 70 years, 88% in women aged 70 to 75 years, and 80% in women older than 75 years (P<.01).

Women receiving hormone replacement (with or without alendronate) had the expected increased incidence of menstrual spotting, cramps, and breast tenderness when compared with women in the 2 groups not receiving hormone replacement (Table 4). Adverse events with combination therapy were similar to those with hormone replacement alone (Table 4). We observed no between group differences in potential adverse effects associated with alendronate (ie, esophagitis, indigestion, heartburn, or dysphagia) (Table 4). The number of hospitalizations, myocardial infarctions, falls, and clinical fractures was similar in all 4 groups. Participants in all groups lost height (range, −0.4 to −1.1 cm), but the differences between groups were not significant.

This is the first study to date to compare the efficacy of hormone replacement, the bisphosphonate alendronate, and combined therapy with both agents in community-dwelling, elderly women. We found that combination therapy yielded greater improvements in BMD at the spine and total hip than did monotherapy with hormone replacement or alendronate. Moreover, at 3 years, the BMD increases observed with combination therapy exceeded those seen in most clinical trials of women with osteoporosis whether treated with alendronate, risedronate, or raloxifene hydrochloride.^24- 28 Furthermore, hormone replacement plus alendronate was safe and well tolerated. Finally, in this head-to-head comparison of alendronate and hormone replacement, we found greater improvements in BMD at the hip, and more responders to therapy with alendronate.

Previous studies with younger women have suggested trends similar to those observed in this trial.^9,29 Bone et al⁹ demonstrated in a group of women, all of whom had had a prior hysterectomy and were a decade younger than our cohort, that combined estrogen and alendronate resulted in greater improvements in BMD at the spine and femoral neck than either agent alone. However, they did not find differences between BMD increases at 2 years in women receiving estrogen vs alendronate. Women were excluded if they had moderate hypertension, recent cardiac disease, dyspepsia, or if they were recently taking aspirin or nonsteroidal anti-inflammatory medications. In contrast, we included such participants. Harris et al³⁰ administered conjugated estrogen with or without risedronate to 524 postmenopausal women for 1 year. Lumbar spine BMD was not found to be significantly different between the 2 groups, but femoral neck BMD in the combination therapy group was approximately 1% greater than in women taking estrogen alone. In our study, not only was the greatest response noted with combination therapy at all sites, but alendronate therapy resulted in a greater increase in BMD at the hip than hormone replacement therapy.

Investigators have reported that more than 95% of women taking alendronate respond to therapy.³¹ Our study confirmed the high response rate that has been observed with alendronate in younger postmenopausal women. However, few data are available in older women receiving hormone replacement or combination therapy. In the Postmenopausal Estrogen/Progestin Intervention trial with younger women (mean age, 56 years), approximately 1.5% of participants receiving hormone replacement therapy had a reduction in spine BMD in the first 12 months (defined as BMD loss of 1% or more in the trial), and 5% to 8% had a reduction in spine BMD from months 12 to 36.¹⁹ At the hip, women showed a reduction in BMD of 14.5% in the first 12 months and of 11.8% from months 12 to 36. To our knowledge, our study is the first long-term clinical trial using standard doses of combined continuous hormone replacement therapy in older women with and without a uterus, and it suggests that over 3 years, 19% of participants had a reduction in BMD at the total hip, and 23% had a reduction in BMD at the trochanter. Only 7% had a reduction in BMD at the posteroanterior lumbar spine (Table 2). However, because women often have atypical calcifications outside of the spine that falsely elevate the posteroanterior lumbar spine measurements, the lateral lumbar spine nonresponse rate of 16% may be a more appropriate reflection of the impact on vertebral bone mass.

Our study has several limitations. First, it was not powered to examine fractures as an outcome. Although the increases of bone mass with combination therapy are greater than with single therapy, use of combination therapy may not translate to greater fracture reduction. Applying a logistic model developed from a meta-analysis of 12 trials modeling the relationship between improvement in vertebral bone mass and reduction in vertebral fractures, combination therapy would provide an additional 10% fracture reduction over hormone replacement and an additional 8% fracture reduction over alendronate.⁸ Both of these translate into clinically relevant benefits. Second, this study may be generalized only to women tolerating hormone replacement for at least the run-in phase. The rationale behind the run-in phase was to minimize withdrawal following randomization. The run-in phase ensured that the 4 groups were of similar size after 3 years, such that women taking hormone replacement could be compared with women taking alendronate, combination therapy, or placebo. Moreover, this study has limited power to examine differences in outcomes of women taking estrogen plus progesterone vs unopposed estrogen. Finally, this study did not have the power to measure differences among treatment groups in the occurrence of rare but serious adverse events, such as myocardial infarction or venous thrombosis. However, data are available from the Women's Health Initiative³ and Heart and Estrogen/Progestin Replacement Study³² that address the risk of complications with hormone therapy. There is no reason to suppose that the risk would differ substantially with the addition of a bisphosphonate to the hormone regimen. The main results of this study (bone mass density with combination therapy) can help answer the question of whether the potential risk of hormone treatment is worth the estimated 8% reduction in fracture risk when combination therapy is used compared with alendronate alone.

This study has several unique features. While the Women's Health Initiative excluded women older than 79 years of age, we included a cohort approximately a decade older than women in the Women's Health Initiative. We included elderly postmenopausal women receiving a standard dose of conjugated estrogen with or without medroxyprogesterone. We also included women with osteopenia or osteoporosis. In addition, we included less-healthy healthy elderly women compared with the majority of other trials for postmenopausal osteoporosis.^{24- 25,27,30- 31,33} Participants with cardiovascular disease, hypertension, gastrointestinal tract disorders, mild renal insufficiency, or those who were taking aspirin or nonsteroidal anti-inflammatory medications were allowed to participate, reflecting typical participants in clinical practice. In addition, the study was performed at 1 center using 1 bone mineral densitometer with daily quality control. Analysis was performed with a conservative mixed-models approach, which incorporated all data available. Similar results (not shown) were found using a last-observation-carried-forward analysis. Furthermore, another advantage of the study is the duration of 3 years. Since the changes in BMD in 1 year are of similar magnitude to the coefficient of variation for the dual-energy X-ray absorptiometry scan used to measure bone mass density, an interval greater than 1 year is needed to assess changes in BMD. Moreover, despite 3 years of therapy in an elderly population, the completion rate was 90%. Finally, this study is the first long-term, head-to-head comparison of hormone replacement and alendronate in elderly women.

In summary, in community-dwelling elderly women, combination therapy with alendronate and hormone replacement is safe and efficacious. Monotherapy with alendronate was shown to be superior to hormone replacement, and combination therapy with both was shown to be superior to either alone. This study provides additional information for the growing body of data on the risks and benefits of hormone replacement therapy.^34- 36 Although findings from the Women's Health Initiative suggest an increased risk when hormone replacement is used for prevention of chronic disease, the study also demonstrated a statistically significant reduction in hip and vertebral fractures. Recently, investigators from the Heart Estrogen/Progestin Replacement Study reported that hormone replacement therapy resulted in a 35% lower risk of type 2 diabetes mellitus during the 4-year trial.^37- 38 More information is needed regarding other potential benefits of hormones or unopposed estrogen, such as improvement in cognitive function, quality of life, mobility, or reduction in depression, falls, and colon cancer in older women.^34,39- 41 However, for elderly women in whom hormone or estrogen replacement is a therapeutic alternative, combination therapy with a bisphosphonate can be considered to further improve skeletal integrity and may be an option for women who fail to achieve an adequate response on monotherapy or for patients with more severe disease for whom monotherapy would be less desirable.