Long-term Effects of Raloxifene on Bone Mineral Density, Bone Turnover, and Serum Lipid Levels in Early Postmenopausal Women:  Three-Year Data From 2 Double-blind, Randomized, Placebo-Controlled Trials FREE

RALOXIFENE hydrochloride is a nonsteroidal benzothiophene that has been classified as a selective estrogen receptor modulator (SERM) based on its demonstrated ability to affect bone and lipid metabolism similarly to estrogen, while simultaneously antagonizing the effects of estrogen in the uterus and breast.^1- 4

Raloxifene given to healthy postmenopausal women for 2 years suppressed bone turnover to the normal premenopausal range; increased bone mineral density (BMD) in the spine, hip, and total body; lowered serum total cholesterol and low-density lipoprotein cholesterol (LDL-C) levels; and did not stimulate proliferation of the endometrium.⁵ Results from the Multiple Outcome of Raloxifene Evaluation (MORE) study indicate that raloxifene treatment reduces new vertebral fracture incidence by about half relative to placebo in older postmenopausal women with osteoporosis treated for 3 years,⁶ a magnitude of effect comparable to that of bisphosphonates.⁷ The potential utility of raloxifene for both prevention and treatment of osteoporosis mandates long-term study. Two clinical trials, each of 5 years' total duration, were identically designed to determine the long-term effects of raloxifene treatment on BMD and lipid levels in healthy postmenopausal women.⁵ Herein we report results from a planned 3-year analysis from these 2 double-blind trials comparing placebo with raloxifene doses of 30, 60, or 150 mg/d.

This analysis includes 36-month data from 1145 women enrolled in 2 double-blind, placebo-controlled clinical trials. Study 1 was conducted in Western Europe (N = 601) and study 2 was conducted in the United States and Canada (N = 544). Both trials enrolled healthy women, aged 45 through 60 years, and between 2 and 8 years (inclusive) beyond their last menstrual period at baseline.

For both studies, systemic hormone replacement therapy was permitted before, but not during, the study, and only if it had been discontinued at least 6 months before entry. Use of vaginal estrogens, except estradiol, was permitted up to an average of 3 times per week during the study. Concomitant use of progestins, androgens, bone-active agents, or other SERMs was not permitted. Complete inclusion and exclusion criteria have been published previously.⁵

A randomized block design was used to assign women equally to receive daily either raloxifene hydrochloride, 30, 60, or 150 mg, or an identical placebo. All women were provided with supplemental elemental calcium (400 to 600 mg/d). Dietary vitamin D was not supplemented. Study visits were scheduled every 3 months for the first 2 years and every 6 months thereafter.

The protocols were approved by local ethical review boards, and all women provided written informed consent for participation in accordance with the principles outlined in the Declaration of Helsinki.

Blood samples were obtained after at least a 6-hour fast every 3 months during the first 2 years, and every 6 months thereafter. Bone turnover was assessed by measurements of serum osteocalcin (ELISA-OSTEO IRMA; CIS Biointernational, Gif-sur-Yvette, France), serum bone-specific alkaline phosphatase (Ostase IRMA; Hybritech, San Diego, Calif), and urine type I collagen C-telopeptide fragments corrected for creatinine excretion (CrossLaps ELISA; Osteometer Biotech A/S, Herlev, Denmark). These analyses were performed centrally (Hôpital Edouard Herriot, Lyon, France, and Covance Laboratories, Indianapolis, Ind, and Geneva, Switzerland). Analyses of serum lipid concentrations and routine laboratory measurements were performed centrally using routine methods (Covance, Indianapolis and Geneva).

Bone mineral density of the spine (L1-L4), hip, and total body (subset of women) was measured twice yearly during the first 2 years and annually thereafter by dual energy x-ray absorptiometry using Hologic QDR-1000, 1500, 2000, or 4500 machines (Hologic, software version 7.10B, Waltham, Mass). Bone mineral density of the nondominant forearm was determined in a subset of women in study 1 by single energy x-ray absorptiometry (Osteometer DTX-100, software version 1.51B; Osteometer Meditech A/S, Roedovre, Denmark) and study 2 (dual energy x-ray absorptiometry, Hologic QDR-1000 and 2000). All scans were checked for quality by a central reader (Quality Assurance Center, Herlev, Denmark, or Osteoporosis Research Group, San Francisco, Calif). Scans for a subset of women in study 1 were reanalyzed to determine BMD in the new region of interest (N-ROI), a region in the ultradistal forearm with a high content of trabecular bone.⁸

At each scheduled visit (no less than every 6 months), women were asked in general terms about the occurrence and severity of adverse events. An adverse event that occurred for the first time or worsened in severity after randomization was considered "treatment-emergent," which does not imply causality. Only treatment-emergent adverse events were analyzed. Women were also asked to grade adverse events as mild, moderate, or severe. If a woman withdrew from the study because of adverse event(s), the predominant adverse event contributing to the withdrawal was specifically captured.

All analyses were performed using an intent-to-treat approach. The data set included all women who had at least 1 follow-up visit after randomization. For women who withdrew from the study before the 3-year visit (referred to as end point), the last available observation was carried forward to subsequent visits. The changes and percentage changes in BMD from baseline to 6, 12, 18, 24, and 36 months were evaluated by analysis of variance, including a term for therapy and study. Initially, a term for therapy-by-study interaction was included and tested for significance at the .10 level. Since this interaction was not significant, it was deleted from all models. For BMD measurements wherein the therapy effect was significant at the .029 level of significance, a pairwise comparison was also performed at the .029 two-sided level of significance, reflecting adjustment for an interim analysis at 2 years.

The changes and percentage changes for biochemical markers of bone turnover and serum lipid levels were not normally distributed (Shapiro-Wilk's test of normality).⁹ Therefore, these data were ranked and analyzed using the analysis of variance described above. Median changes were used to represent the central tendency, and SEs for median changes were estimated using the d-delete jackknife method.¹⁰ These measures were tested at the .05 two-sided level of signifi-cance for overall therapy effect and for each pairwise comparison.

For measures determined to be normally distributed, means and SDs are presented. For those that were not normally distributed, medians and jackknife SEs are presented.

Safety measurements having continuous distribution (eg, laboratory results, vital signs) were analyzed in a manner similar to efficacy measurements. Study discontinuations and safety assessments having binary responses (eg, adverse events) were analyzed using the Cochran-Mantel-Haenszel tests for general association stratified by study.

The randomization procedure and identical inclusion and exclusion criteria contributed to similar study groups being enrolled in the 2 trials, as evidenced by the baseline characteristics of enrolled women (Table 1). Treatment responses for key end points were also comparable between studies, and there was no difference between studies in the proportion of women who had withdrawn after 3 years. Throughout the follow-up interval, about 90% of participants in each treatment group were compliant with study medication, defined as taking at least 80% of dispensed tablets (data not shown).

Bone mineral density decreased by 1% to 2% from baseline at all body sites (and by 3% at the wrist) in the placebo group. In the raloxifene groups, there was an increase in spine and total hip BMD during the first 2 years, which was maintained during the third year (Figure 1). The response to raloxifene at the femoral neck was similar to the total hip (data not shown). The BMD responses to raloxifene, compared with placebo, were similar across anatomic sites (Figure 1 and Table 2). There were no statistically significant differences in the mean percentage changes in spine and hip BMD among raloxifene groups at end point, although the response to 30-mg raloxifene tended to be less than that of the 2 higher doses. Total body BMD was increased at 3 years by an average of 1.7% with raloxifene therapy compared with placebo (Table 2). Raloxifene, 60 and 150 mg, significantly attenuated, but did not completely prevent, BMD decline at the ultradistal radius or the forearm new region of interest (Table 2).

Raloxifene therapy significantly decreased bone turnover to premenopausal levels, as reflected by decreases of serum osteocalcin, serum bone-specific alkaline phosphatase, and urine CrossLaps/creatinine^11- 12 (Table 3). The placebo group also had a slight, but significant, decrease in most markers (Table 3). The responses of serum CrossLaps, serum N-mid osteocalcin, and urine and serum type I collagen fragment N-telopeptide were similar to those of the other bone turnover markers (Table 3). Raloxifene treatment significantly decreased serum total and LDL-C levels (Table 4) throughout 3 years, whereas high-density lipoprotein cholesterol (HDL-C) and triglyceride concentrations were unchanged from baseline after 3 years of raloxifene treatment.

During 3 years of treatment, the tolerability of raloxifene was comparable to that of placebo. Overall, 37% of women withdrew from the study at or before the 3-year visit. The study discontinuation rates were similar across treatment groups, and there were no treatment differences in the reasons for early study withdrawal. A total of 164 (14%) women withdrew due to an adverse event, the most common reason for study withdrawal.

At least 1 adverse event was reported by 88% of the participants, with no difference among treatment groups overall. The only adverse event for which there was both a statistically significant treatment difference and a higher reported incidence among raloxifene-treated women was hot flashes (25% in the 60-mg raloxifene group vs 18% in the placebo group) (Table 5). Hot flashes were generally mild and did not increase study withdrawals (1.7% for 60-mg raloxifene vs 2.4% for placebo). Symptoms usually accompanying hot flashes, such as insomnia and sweating, were not reported more frequently by raloxifene-treated women (data not shown). Other adverse events that have been associated with menopause, therapy with other SERMs, estrogens, or estrogen-progestin replacement therapies were also not reported more frequently during raloxifene therapy compared with placebo (Table 5).

During 3 years of treatment, raloxifene, 30 to 150 mg/d, reduced bone turnover to within normal premenopausal ranges^11- 12 and preserved BMD above baseline at the spine, hip, and total body, supporting the use of 60 mg/d of raloxifene for long-term prevention of postmenopausal bone loss. Small decreases in markers of bone turnover in the placebo group, possibly attributable to calcium supplementation, were not sufficient to prevent loss of BMD.

Data from 2 trials were integrated for this study. The validity of data integration was evidenced by the comparable baseline characteristics and responses to therapy between the 2 studies. Throughout the follow-up period, the random assignment to therapy and the double-blind were maintained; medication compliance was as expected and was unaffected by drug assignment; and attrition from the study was as expected for an asymptomatic disease prevention study and was unaffected by treatment.

The similarity of BMD responses to raloxifene in different body regions that contain either primarily trabecular (ie, spine) or cortical bone (ie, femoral neck) suggests that the effect of raloxifene is similar on these 2 bone types. This is in contrast to other antiresorptive treatment modalities, for example, bisphosphonates, which preferentially increase BMD in the spine.^13- 15 One hypothesis accounting for the differential effect of bisphosphonates is based on an increase in the amount of mineral per volumetric unit of bone. Animal and human bone microradiography studies indicate that, when bone metabolism is markedly reduced by bisphosphonates, the second phase of mineralization (which accounts for about 30% of mineral deposited into bone) is significantly prolonged, thus increasing the amount of mineral per unit of bone.^16- 17 Such an effect is more likely to be observed in sites rich in cancellous bone, such as the lumbar spine, characterized by higher turnover rates than cortical bone. Thus, the relatively large increase in BMD observed with antiresorptive agents that markedly reduce bone turnover (below the normal premenopausal range) could be due initially to filling of the remodeling space (greatest in cancellous bone) and later to an increase in bone mineral per unit volume of bone (also greatest in cancellous bone). Such a hypothesis is consistent with the difference in bone turnover marker suppression between raloxifene (about 30%) and potent bisphosphonates (>50%).^14- 15

A similar BMD response to raloxifene in the prevention studies was observed in an osteoporosis treatment trial, Multiple Outcomes of Raloxifene Evaluation (MORE), which enrolled osteoporotic women who were about 10 years older at baseline⁶ than subjects in the present report. Although these BMD increases seem modest, patients in MORE treated with raloxifene experienced a significant reduction of risk for incident vertebral fracture at 3 years of therapy.⁶ Thus, although the BMD response to raloxifene in the spine is substantially less than that observed for bisphosphonates,^13- 15 the reduction in vertebral fracture risk appears to be similar. These and other observations suggest that on-treatment change in BMD does not accurately predict clinical outcomes in response to antiresorptive therapies. Whether bone turnover marker responses will prove to be better predictors of fracture prevention efficacy for antiresorptive therapies remains to be determined.

The raloxifene-induced reductions in LDL-C without change in HDL-C or triglyceride concentrations are consistent with previous results,^5,18 and indicate that these effects are sustained long-term. The responses in markers of cardiovascular risk to raloxifene treatment differ from those to oral hormone replacement therapies. In a 6-month study comparing raloxifene with continuous-combined conjugated equine estrogens and medroxyprogesterone acetate (HRT), there was a similar effect of the therapies on LDL-C and total cholesterol, whereas raloxifene had a lesser effect on lipoprotein(a) and no effect on HDL-C.¹⁸ However, raloxifene did not affect serum triglyceride concentrations, which were elevated by HRT. While HRT enhanced fibrinolysis, raloxifene had no effect on fibrinolysis but markedly reduced serum fibrinogen concentration.¹⁸ Although surrogate markers of cardiovascular risk are interesting surrogate end points, only large-scale controlled clinical trials can provide conclusive evidence of coronary heart disease benefit. The recently published Heart and Estrogen/progestin Replacement Study (HERS),¹⁹ which enrolled women with established coronary disease, found no benefit of HRT compared with placebo for nonfatal myocardial infarction or coronary heart disease death. The effects of raloxifene therapy on fatal and nonfatal cardiovascular events are among the primary end points in the Raloxifene Use for The Heart (RUTH) trial,²⁰ a multicenter, randomized double-blind trial of raloxifene, 60 mg/d, compared with placebo in 10,000 postmenopausal women who have risk factors for or history of heart disease.

The overall tolerability profile of raloxifene in these studies was similar to that of placebo, although a small increase in incident hot flashes was found (excess risk, appoximately 7 of 100 treated). Hot flashes were typically reported as mild, were not associated with other symptoms, and did not contribute to study withdrawal. Importantly, there was no evidence that raloxifene aggravated symptoms or signs of menopause such as vaginitis, and raloxifene therapy was also not associated with symptoms commonly reported during therapy with other SERMs (eg, leukorrhea)²¹ and HRT (eg, breast pain, vaginal bleeding). Although reported rarely in these osteoporosis prevention studies, raloxifene therapy has been associated with development of venous thromboembolism in older women,²² as have HRT¹⁹ and tamoxifen therapy.²³ Raloxifene therapy has also been shown to reduce the risk of invasive breast cancer (a secondary end point in an osteoporosis treatment trial) by 76% in women with osteoporosis at a median of 40 months of observation.²⁴ As with venous thromboembolism, breast cancer was reported rarely in these osteoporosis prevention studies.

In conclusion, raloxifene therapy in healthy, relatively young postmenopausal women resulted in continuous and effective skeletal antiresorptive effects, with decreases in markers of bone turnover and preservation of BMD during 3 years. Raloxifene therapy also reduced serum LDL-C level and was well tolerated. For these reasons, raloxifene should be considered as an adjunct to exercise and nutritional optimization for long-term osteoporosis prevention in healthy postmenopausal women with risk factors for osteoporosis. Preliminary results showing favorable effects of raloxifene on the risk of breast cancer and the favorable effects on markers of cardiovascular risk suggest that the indications for raloxifene therapy may be broadened in the future.²⁰

Accepted for publication June 16, 2000.

Support was provided by Eli Lilly and Company in the form of research grants to the individual study sites.

The following were principal investigators: Austria: J. Huber, MD (Vienna); Belgium: J. P. DeVogelaer, MD (Brussels); Denmark: H. Lawaetz, MD (Ballerup); France: P. D. Delmas, MD (Lyon); Germany: J. Semler, MD (Berlin), J. Happ, MD (Frankfurt); Italy: M. L. Brandi, MD (Florence), C. Gennari, MD (Siena); the Netherlands: S. E. Papapoulos, MD (Leiden), J. I. Puyenbroek, MD (Amsterdam); United Kingdom: I. Smith, MD, and S. Young, MD (Lancashire, England); United States: S. R. Weiss, MD (San Diego, Calif), C. Arnaud, MD (San Francisco, Calif), H. H. McIlwain, MD (Tampa, Fla), C. Johnston, MD (Indianapolis, Ind), T. J. Weber, MD (Durham, NC), R. Lindsay, MD (West Haverstraw, NY), S. S. Miller, MD (San Antonio, Tex), H. Heath III, MD, and T. P. Knecht, MD (Salt Lake City, Utah).

We thank Patrick Garnero, Lyon, France, for bone marker analyses; Anette Møllgaard for assistance in forearm bone mass reanalysis; colleagues from Osteometer Biotech for assistance in measurement of serum CrossLaps and N-Mid osteocalcin; Steven D. Watts for statistical programming; and Erin Walls, ELS, for editorial assistance. We also gratefully acknowledge the enthusiastic, steady participation of the women volunteers, and the hundreds of skilled and educated research staff.

Reprints: C. Conrad Johnston, Jr, MD, Indiana University School of Medicine, Emerson Hall Room 421, 545 Barnhill Dr, Indianapolis, IN 46202.