For the majority of pharmacological interventions used for osteoporosis, the duration of clinical trials recommended by regulatory agencies is 3 years; hence, robust data are limited to this period of time. However, many patients remain at increased risk of fracture after this duration of therapy, raising questions about both the efficacy and safety of longer periods of treatment.1 Because of the unique pharmacokinetics of bisphosphonates and, in particular, their long retention time in bone, it is possible that beneficial effects on fracture risk may persist for some time after treatment is stopped. The possible association between bisphosphonate therapy and two rare but serious conditions, namely osteonecrosis of the jaw (ONJ) and atypical femoral fractures, adds another dimension to this discussion. Some studies have linked these two rare adverse events to length of therapy.2–4 The long residence time in bone of bisphosphonates, along with possible time-related adverse events, have led to the concept of a drug holiday. The drug holiday is a period of time, after continuous bisphosphonate therapy, when treatment is stopped.5 As suggested by the term “holiday,” treatment is restarted at some point. Along with the topical subject of the drug holiday is the equally important issue that if treatment is stopped for any significant period of time, the patient might be exposed to an increased risk of fracture.
Until recently, information on the effects of longer-term treatment and its discontinuation was restricted to alendronate and risedronate. In an extension of the Fracture Intervention Trial (Fracture Intervention Trial Long-term Extension [FLEX]), postmenopausal women who had been treated with alendronate for a mean of 5 years were randomized to treatment with either 5 or 10 mg daily of alendronate or placebo for an additional 5 years.6 In the group assigned to placebo, bone mineral density (BMD) in the hip declined substantially although it remained above pretreatment values at the end of the study. Biochemical markers of bone turnover increased modestly. The incidence of all clinical fractures and of nonvertebral fractures was similar in the continuation and discontinuation groups, but the risk of clinical (but not morphometric) vertebral fractures was significantly lower in those who continued alendronate therapy. However, reduction in nonvertebral fractures with bisphosphonate therapy has generally been demonstrated only in women with osteoporosis,6–8 and because many of the women in FLEX did not have osteoporosis, the power to show an effect would likely be reduced. Post hoc subgroup analysis of the FLEX study indicated that the risk of both nonvertebral and clinical vertebral fractures increased with lower baseline BMD or prevalent vertebral fracture. Subsequently, another post hoc subgroup analysis from the FLEX study suggested that continuation of alendronate was associated with significantly lower risk of nonvertebral fractures in women without a vertebral fracture at baseline but with a baseline femoral BMD T-score ≤ −2.5.9 Although due caution about the interpretation of post hoc analyses is in order, together the data give reason to be concerned that global fracture protection afforded by this bisphosphonate may not persist if treatment is discontinued for a prolonged period of time.
Fewer data are available for risedronate. In postmenopausal women treated with risedronate or placebo for 3 years, assessment of BMD and bone turnover markers was performed 1 year after discontinuation of treatment. In the group previously treated with risedronate, spine and hip BMD decreased significantly although remaining above baseline values and bone turnover markers increased to placebo levels. At the end of the 1-year extension period, the incidence of morphometric vertebral fractures remained significantly lower in the group that had previously received risedronate.10 Recently, the effects of 1 year's discontinuation after either 2 or 7 years of risedronate have been reported. Total hip and trochanter BMD, but not lumbar spine or femoral neck BMD, decreased during the year off treatment in both groups and urinary cross-linked N-telopeptide of type I collagen (NTX) increased toward the baseline.11
In this issue of Journal of Bone and Mineral Research, Black and colleagues report the results of a 3-year extension of the Health Outcomes and Reduced Incidence with Zoledronic Acid Once Yearly (HORIZON) pivotal fracture trial of zoledronic acid in postmenopausal women with osteoporosis.12 A subset of women who had participated in the treatment arm of the core trial were randomized to continue zoledronic acid, 5 mg by intravenous infusion once yearly, or to receive intravenous infusion of a placebo. Femoral neck BMD, the primary end-point, remained constant in the group that continued to take zoledronic acid and showed a small decrease in the placebo group. Increases in femoral neck BMD over the 6-year period were 4.5% and 3.1%, respectively, with only a small, albeit statistically significant, difference between the continuation and noncontinuation groups at the end of the study (mean, 1.04%; 95% confidence interval, 0.4%–1.7%). BMD at other sites showed a similar pattern and serum N-terminal propeptide of type I collagen (PINP) showed a small increase in both groups during the 3-year extension period (19% versus 33% in the treatment and placebo groups, respectively). There were no significant differences between the groups in nonvertebral, clinical vertebral, hip, or all clinical fractures. However, significantly fewer morphometric vertebral fractures occurred in the women who continued to take zoledronic acid than in those in the placebo group (14 versus 30). This difference remained significant when the analysis was restricted to moderate or severe vertebral fractures.
With respect to long-term safety, no specific issues were identified in the studies described above. In particular, although recent evidence supports an association between the duration of bisphosphonate therapy and the development, albeit rarely, both of ONJ and atypical fractures,13–15 ONJ was reported in only one woman in the zoledronic acid extension trial and atypical fractures were not seen in any of the studies. Nevertheless, these remain potential concerns with long-term treatment and emphasize the need for regular review of the need to continue treatment.
How can these studies be used to inform clinical practice? The first lesson is that there are differences between individual bisphosphonates, and recommendations based on the available data should be drug-specific. Although no head-to-head studies have been conducted, the offset of effect on BMD and bone turnover markers appears to be most rapid for risedronate, intermediate for alendronate, and slowest for zoledronic acid. Second, for zoledronic acid, beneficial effects on BMD may persist for longer than 3 years after discontinuation. Third, in the case of alendronate and zoledronic acid, bone turnover shows some recovery after discontinuation but remains suppressed for several years, whereas the return toward pretreatment bone turnover is much more rapid after discontinuation of risedronate. Fourth, although none of the studies was powered to demonstrate effects on fracture rate, the results of FLEX and the HORIZON extension study indicate that in women at high risk of fracture, discontinuation of alendronate or zoledronic acid for 3 or more years may not be advisable. Global fracture protection is not sustained. In particular, women who develop one or more morphometric vertebral fractures during therapy and who have a femoral neck or total hip T-score ≤ −2.5 may be at the greatest risk of developing new vertebral fractures if therapy is discontinued.6, 16 The effect of risedronate therapy on vertebral fracture risk appears to be maintained in the first year after withdrawal,10 but the return of bone markers to or toward placebo levels and the accompanying bone loss during this period suggest that treatment benefits may be lost relatively soon thereafter.
In deciding the optimal duration of therapy, the benefits and risks of both continuation and discontinuation must be considered. Long-term treatment is associated with fracture reduction but may increase the risk of rare adverse effects such as ONJ and atypical fractures. Discontinuation might reduce the risk of ONJ and atypical fractures but may also be associated with reduced protection against fracture. For alendronate and risedronate, an initial 5-year period of treatment seems reasonable, with assessment of the need for further treatment at the end of that time based on fracture history and BMD. If a drug holiday is advised, reassessment of risk should be performed after 1 to 2 years for alendronate, but earlier, perhaps at 1 year, for risedronate. In most patients treated with zoledronic acid, an initial treatment period of 3 years appears to be sufficient, with reassessment of the need for further treatment after 2 to 3 years. Since the effects of a single infusion of zoledronic acid on BMD and bone turnover markers are maintained for up to 3 years, less-frequent dosing than the approved once-yearly regimen may be possible, although the effect of this approach on fracture rate has not been established.17 Although the HORIZON extension study adds useful information to previous studies of discontinuation of other bisphosphonates, at present there is no strong evidence base to guide clinicians in making decisions about duration of treatment and drug holidays. Nevertheless, the strength of evidence for fracture reduction in high-risk patients and the rarity of long-term adverse effects indicate that in the majority of individuals the benefits of continued treatment outweigh the risks and that treatment should be continued long-term in individuals who remain at high risk of fracture.