Fracture risk remains reduced one year after discontinuation of risedronate
- First Online:
- Cite this article as:
- Watts, N.B., Chines, A., Olszynski, W.P. et al. Osteoporos Int (2008) 19: 365. doi:10.1007/s00198-007-0460-7
- 656 Views
One year after discontinuation of three year’s treatment with risedronate, BMD decreased at the lumbar spine and femoral neck and bone turnover markers returned to control group levels. Despite these changes, the risk of new morphometric vertebral fractures remained lower in previous risedronate patients compared with previous control patients.
Differences in bisphosphonate pharmacology and pharmacokinetics could influence persistence or resolution of the effects once treatment is stopped. We investigated changes in intermediate markers—bone mineral density (BMD) and bone turnover markers (BTM)—and fracture risk after discontinuation of treatment with risedronate.
Patients who received risedronate 5 mg daily (N = 398) or placebo (N = 361) during the VERT-NA study stopped therapy per protocol after 3 years but continued taking vitamin D (if levels at study entry were low) and calcium and were reassessed one year later.
In the year off treatment, spine BMD decreased significantly, but remained higher than baseline (p ≤ 0.001) and placebo (p < 0.001), with similar findings at the femoral neck and trochanter. Urinary NTX and bone-specific alkaline phosphatase, which decreased significantly with treatment, were not significantly different from placebo after 1 year off treatment. Despite the changes in intermediate markers, the incidence of new morphometric vertebral fractures was 46% lower in the former risedronate group compared with the former placebo group (RR 0.54 [95% CI, 0.34, 0.86, p = 0.009]).
Despite the apparent resolution of effect on BMD and BTM, the risk reduction of new vertebral fractures remained in the year after treatment with risedronate was stopped.
KeywordsBone mineral densityBone turnoverOsteoporotic fracturePostmenopausal osteoporosisRisedronateVertebral fractures
Osteoporosis is defined as a skeletal disorder characterized by compromised bone strength predisposing to an increased risk of fracture . Bone strength is determined by a combination of structural and material properties of bone, which in turn are influenced by bone turnover . Risedronate, a pyridinyl bisphosphonate, has been shown to affect multiple factors that contribute to bone strength. It decreases bone turnover markers (BTMs) as early as one week after initiation of therapy , increases bone mineralization  and preserves microarchitecture , all of which contribute to increases in bone mineral density (BMD). The contribution of changes in BMD to fracture risk reduction in response to treatment is small. In the risedronate trials, only 11%–18% of the observed reduction in vertebral fracture risk was explained by BMD changes . Non-vertebral fracture risk reduction observed with risedronate appears to be explained to an even lesser extent by BMD changes . Change in BTM may explain as much as 67% of the observed reduction in vertebral fracture risk, but the relationship between change both intermediate markers (BMD and BTM) and fracture risk reduction appears to be non-linear . For BMD changes, the vertebral fracture risk reduction observed over three years of risedronate treatment is comparable whether BMD increases between 0–5% or more than 5% . Additional data suggest that the fracture risk reduction of bisphosphonate-treated patients compared with placebo is similar whether patients gain BMD or lose BMD [9, 10]. Although change in BTM may explain more of the reduction in vertebral fracture than change in BMD, at last half of fracture risk reduction is unexplained by changes in BTM [8, 11]. It is therefore likely that both the significant rapid reduction in the risk of vertebral and non-vertebral fractures within 6 months of risedronate treatment [12, 13] as well as the sustained reduction in vertebral and non-vertebral fracture risk over 5 years  is only partially explained by changes in BMD and BTM.
Bisphosphonates are unique pharmacological agents. Due to their high affinity for hydroxyapatite of bone and their relatively slow release from the skeleton, they have the potential of yielding effects on bone extending significantly beyond the duration of the actual treatment . This poses a need for patient monitoring after discontinuation, and BMD and bone turnover markers have been proposed for this purpose. The lack of association between these intermediate parameters and fracture reduction while patients are on treatment, however, makes it difficult to interpret changes in these parameters after discontinuation of treatment.
To address the resolution of effects of risedronate therapy, we studied postmenopausal women with osteoporosis who completed a 3-year, double-blind treatment period in which they received risedronate 5 mg daily or placebo and were then followed for an additional year without risedronate treatment.
Study design and subjects
Women (N = 2,458) who were at least 5 years postmenopausal, younger than 85 years and had either two or more vertebral fractures or one vertebral fracture and low spinal bone mineral density (T-score ≤−2) were enrolled in the original double-blind, placebo-controlled study and were randomly assigned to receive risedronate 2.5 or 5 mg or placebo daily for 3 years. (The risedronate 2.5 mg group was discontinued early by protocol amendment.) All women also were supplied with 1000 mg of calcium daily and, if baseline serum 25-hydroxyvitamin D levels were low, received vitamin D supplementation. A full report of this study has been published . Except for 121 women who had bone biopsies at the end of year three , all women who completed the 3-year treatment period were given the option of remaining in the study for the fourth year, where active treatment and placebo were discontinued but calcium and vitamin D supplementation was continued as before. Patients and clinical personnel remained blinded to the original treatment assignments throughout the 4 years of observation.
BMD of the lumbar spine and proximal femur were obtained during the 3 years of treatment and at the end of the open-label follow-up period using methods previously described . Serum bone-specific alkaline phosphatase (BAP) was measured periodically in patients enrolled at a subset of study centers during both the treatment and follow-up periods using the Tandem R-Ostase immunoradiometric assay (Hybritech Inc, San Diego, California, USA). Second morning void urine samples for measurements of bone resorption markers (deoxypyridinoline crosslinks) were obtained and measured periodically. For the purposes of this study, however, to avoid artifacts caused by batch to batch variation, cross-linked N-telopeptides of type I collagen (NTX) in stored samples obtained in the extension study were determined together with samples obtained at the baseline of the original study in a single batch by automated analyzer (Vitros ECi; Ortho Clinical Inc., Rochester, NY, USA) (interassay CV 6.7%) as previously described . All samples were continuously stored at −80°C until analysis. Measurements were corrected for creatinine using a dry chemistry method.
In the initial three-year study, new vertebral fractures were determined from lateral spine radiographs using morphometric and semiquantitative methods; differences between the two were adjudicated, as previously described . For the purposes of the extension study reported here, the radiograph taken at the end of the 4th year was compared with the one at year three, at a central site, by radiologists who were unaware of the subjects’ previous treatment assignment.
Adverse event reports were collected at study center visits, which occurred every 3 months during the 3-year treatment period and after 6 and 12 months during the follow-up period. Non-vertebral fractures regardless of location or cause were reported as adverse events. During the original three-year study, radiographic confirmation was required. During the extension period, radiographs to confirm the non-vertebral fractures were not required but were available in all but six cases. Physical examinations were performed at baseline, at the end of the 3-year treatment period and at the end of the follow-up period. Standard serum chemistry and hematology tests were performed on samples obtained at baseline and at intervals during the treatment period and at the beginning and end of the follow-up period.
Analyses of the follow-up data included all women who completed the 3-year treatment period and entered follow-up. Missing values at month 48 were imputed using the last observation carried forward method. Characteristics of the population at baseline of the original study and at baseline of the extension study were summarized for two groups of women according to whether they received placebo or risedronate 5 mg in the 3-year treatment period. The primary efficacy analysis compared the mean percent change from the original baseline in lumbar spine BMD between month 48 and month 36 for each treatment group separately. Other pre-specified analyses included the within-group comparisons about the mean percent change from baseline in femoral neck BMD and BTMs, the between-group comparisons about the percent change from baseline in BMD and BTM, and the incidence of new vertebral fracture during follow up period. Within-group comparisons were made using paired T-tests for BMD measurements and using Wilcoxon Signed Rank tests for BTM measurements. Between-group comparisons were made using analysis of variance model with treatment groups, pooled center and stratum as factors for BMD changes; Wilcoxon Rank Sum tests for BTM changes. The incidence of new vertebral fracture was compared between treatment groups using the Cochran–Mantel–Haenszel (CMH) test, stratified by the number of vertebral fractures at month 36 (3 groups: no vertebral fracture, one vertebral fracture and two or more vertebral fractures). The CMH estimator of the relative risk of new vertebral fracture and its 95% confidence interval was also calculated. All statistical tests were performed at the 0.05 level of significance.
Baseline characteristics of the women enrolled in the calcium follow-up period at the beginning of the initial trial
Placebo N = 361
Risedronate 5 mg N = 398
Age – years
Time since menopause – years
Weight – kg
Height – cm
Lumbar spine bone mineral density – T-score
NTX/Cr – nmol BCE/mmol (median)
BAP (μg/L) (median)
No. of vertebral fractures - mean (SE)
Characteristics of the women enrolled in the calcium follow-up period at the entry of the follow up trial
Placebo N = 361
Risedronate 5 mg N = 398
Age – years
Time since menopause - years
Height – cm
Lumbar spine bone mineral density T-score
NTX/Cr - nmol BCE/mmol (median)
BAP (μg/L) (median)
No. of vertebral fractures - mean (SE)
In the year off of treatment, BMD of the femoral neck declined in the former risedronate 5 mg group (−1.23%; 95% CI, −1.87% to −0.60%; p < 0.001). A decrease was also observed at the trochanter after discontinuing risedronate (−1.57%; 95% CI, −2.19% to −0.94%; p < 0.001), but this value remained significantly higher than baseline (p < 0.001). The values in both the femoral neck and trochanter in the risedronate group remained significantly higher than placebo at the end of year 4. The difference between former risedronate and placebo treatment groups at the end of the extension year was 2.32% (95% CI, 1.40% to 3.25%, p < 0.001) at the femoral neck and 3.08% (95% CI, 2.06% to 4.10%, p < 0.001) at the trochanter (Fig. 2b).
BAP percent change from baseline values was available for a subset of 187 women (placebo, 85; former risedronate 5 mg, 102). After stopping treatment, BAP also increased significantly, back to pretreatment levels, and was not different from placebo (Fig. 3b).
In the year following discontinuation of treatment, the adverse event profile in the group previously treated with risedronate was similar to that in the previous placebo group. Hematology, kidney and liver function and serum chemistry findings were unremarkable. Non-vertebral fractures were collected as adverse events and did not require radiological confirmation, although this was obtained in the majority of cases. The number of patients with non-vertebral fractures, regardless of cause and location were 18/361 (5.0%) in the previous placebo group, among them four patients with wrist, three with radius (other than wrist), two with hip fractures) and 19/398 (4.8%) in the previous risedronate group, among them five patients with wrist, four with radius (other than wrist) and two with hip fractures.
In untreated postmenopausal women, low BMD and increased BTM are good predictors of fracture risk, so it is logical that changes in these easily measurable endpoints might serve as potential surrogates for treatment-related fracture risk reduction, the relevant end-point for treatment. Studies based on individual patient data indicate that changes in BMD explain less than 20% [6, 8, 18, 19] and changes in BTM explain approximately 50% of the observed fracture risk reduction associated with antiresorptive treatment. In addition to the limited contribution to explain fracture risk reduction, the relationship between these intermediate end-points and fracture risk reduction appears to be non-linear [6, 8].
Resolution of effect is a topic of little interest in disease areas such as hypertension or hypercholesterolemia, or for biologic agents used for treatment of osteoporosis, as the effect of treatment is assumed to disappear after the drug is stopped. Bisphosphonates are incorporated into bone and have the potential for a residual effect after discontinuation. If there is a residual effect, it might be possible to offer “drug holidays.” Studies reporting the 7-year experience with risedronate  and 10-year experience with alendronate [21–23] suggest that long-term treatment with these agents appears to be safe, with no increased risk of fracture or other adverse effects with the doses used to treat osteoporosis. Adverse treatment effects, such as osteonecrosis of the jaw [24, 25] or over-suppression of bone turnover , may be rare consequences of oral bisphosphonate therapy, but if they occur, a more rapid resolution of antiresorptive effect might facilitate patient management.
During the treatment phases of most osteoporosis treatment studies, all patients receive calcium and many receive vitamin D, resulting in approximately 30% reduction in BTM compared with baseline, and in small but often significant increases in BMD. For clear interpretation of the resolution of a treatment effect, it is therefore necessary to follow both subjects from the former treated and the former placebo group. Studies on discontinuation of treatment with alendronate suggest at least a partial resolution of effect on BMD and BTM [21, 22, 27, 28].
In this study we were able to investigate the resolution of treatment effect on the intermediate end-points BMD and BTM as well as on fracture risk reduction in a previously treated group (risedronate 5 mg daily plus calcium and vitamin D) with a control group (placebo plus calcium and vitamin D). In patients treated for three years with risedronate, BMD decreased significantly in the first year off of treatment, but remained higher than the control group, suggesting a partial resolution of effect. BTM showed a complete resolution of effect to levels of the control group. This is in contrast to the effect observed after 2 years of alendronate, where urine NTX did not return to the level in the control group in the first year off treatment [27, 28].
Because increases in BMD and decreases in BTM have been assumed to be associated with fracture risk reduction on treatment, the maintenance of those effects after discontinuation could be interpreted as an indication for continued anti-fracture efficacy, whereas the resolution of effect on those parameters might be interpreted as a loss of anti-fracture efficacy. It is necessary to have a simultaneous control group to evaluate anti-fracture efficacy of continued or past treatment. Because the existing studies addressing resolution of treatment effect do not have such a control group, inferences regarding anti-fracture efficacy cannot be made. However, in this study, we were able to assess fracture rates and demonstrate a 46% lower risk of vertebral factures in the year after discontinuation of treatment with risedronate 5 mg compared with the control group, similar to what was seen in the original 3-year placebo-controlled trial . Thus, in this study, the partial or complete resolution of antiresorptive effect as measured by BMD and BTM did not indicate loss of anti-fracture effect over the year of observation.
What could explain this apparent discrepancy? The current NIH definition of osteoporosis describes the disease as a “skeletal disorder characterized by compromised bone strength predisposing to an increased risk of fracture” . Clinicians currently measure BMD and BTM because those measurements are readily available and have a logical connection with fracture risk. In the setting of treatment withdrawal, changes in BMD and bone turnover markers do not seem to explain the observed fracture risk reduction. Bone strength is determined by a number of structural properties, including microarchitecture, geometry and material properties such as degree of mineralization and collagen quality, all of which are influenced by bone turnover . Preservation of microarchitecture over three years of treatment [5, 17] and increase of mineralization to the levels of premenopausal women , which have been reported as a result of risedronate treatment, may contribute to the observed anti-fracture effects. It is also theoretically possible that the fracture protective effect of risedronate observed reflected persistent efficacy for the first few months after stopping therapy, and that the salutary effect has dissipated by the end of 12 months following discontinuation. The observed reduction in fracture risk in the year off treatment, is similar in magnitude to what was observed in the first three years on treatment and similar to what was observed in year 4 and 5 of the VERT-MN extension study, where patients were continued on risedronate treatment and compared with placebo , which makes this explanation not very likely.
What are the clinical implications of these findings? There have been recommendations to discontinue bisphosphonate treatment after 3 to 5 years and then follow BMD or BTM to decide when to reinitiate treatment . Our data suggest that the association of changes in these intermediate markers to the observed fracture risk reduction is not clear enough to justify such a strategy for patients treated with risedronate. There are no data for or against such an approach for other bisphosphonates, and until such a relationship has been clearly demonstrated, it should not be assumed to exist.
One of the major strengths of this study is that it includes a control group in post-treatment follow-up with a consistent assessment of vertebral fractures and bone resorption markers analyzed in a single batch. The study also has limitations. By virtue of the fact that is an extension study, not all the patients initially randomized to the original study entered the extension period; however, at baseline, the characteristics of the extension cohort were similar to those of the original cohort. As expected, the group receiving treatment showed increases in BMD and reductions in BTM, but at the start of the extension period, other key characteristics (age, years since menopause and in particular, prevalent vertebral fractures) were nonetheless similar between the two groups. We did not observe a difference in the incidence of non-vertebral fractures in the extension period; however, the number of non-vertebral fractures in the extension period was low, so no clear conclusion can be drawn, whether this is an indication of a gradual loss of effect or rather a chance finding. Other limitations of the study lie in the design — it would have been helpful to have data on BMD and markers of bone turnover at an intermediate timepoint in the 12 month observation period, and it would have been interesting to compare the findings in the group that discontinued risedronate to a group of patients continuously treated with risedronate over the full 4 year period. In the extension period of the multinational arm of the VERT studies, in which patients were treated over five years and compared with a placebo group, the fracture risk reduction of treated versus placebo patients observed in the year 4–5 extension period was similar to that observed in years 1–3 of the Multinational Study and to what we observed in the year off treatment (year 4) of the North American study reported here .
In conclusion, this study demonstrates a significant reduction in vertebral fractures in the year after discontinuation following three years of risedronate treatment, compared with patients who had not been treated, despite partial resolution of the effects of treatment on BMD and a complete resolution of the effects on BTM. These findings are reassuring regarding the persistence of an anti-fracture effect after stopping therapy with risedronate and raise significant questions about the value of intermediate endpoints such as BMD and BTM in guiding clinical decisions after discontinuation of treatment.
We would like to acknowledge Rosemary Hannon, Ph.D. for performing the NTX measurements reported in this study, Ruby Xia and John Banner, M.S. for performing part of the statistical analyses reported in this study as well as all the investigators and staff at the study centers.
Conflict of interest statements
Dr. Watts receives honoraria for lectures from Amgen, Novartis, Procter & Gamble and sanofi-aventis; consulting fees from Amgen, Eli Lilly, Kyphon, Novartis, Procter & Gamble, Roche and sanofi-aventis; and research support through Amgen, Eli Lilly, Novartis, Procter & Gamble and Solvay.
Dr. Chines is a former employee of Procter & Gamble and owns Procter & Gamble shares. He is currently an employee of Wyeth.
Dr. McKeever is a consultant for Procter & Gamble, Merck and Roche and receives research grants from Procter & Gamble, Merck, Lilly and Roche.
Dr. McClung receives research grants from Amgen, Eli Lilly, Merck, Novartis, Procter & Gamble, and Roche. He receives consulting fees from Amgen, Eli Lilly, Merck, Novartis, Procter & Gamble, Roche and sanofi-aventis. He is a member of speaker’s bureaus for Eli Lilly, Merck, Procter & Gamble, and sanofi-aventis.
Dr. Zhou and Dr. Grauer are full time employees of Procter & Gamble and own stocks of the company.