Bisphosphonate drug holiday: who, when and how long

Abstract

Bisphosphonates have been widely used in the treatment of osteoporosis with robust data from numerous placebo-controlled trials demonstrating efficacy in fracture risk reduction over 3–5 years of treatment. Although bisphosphonates are generally safe and well tolerated, concerns have emerged about adverse effects related to long-term use. For most patients with osteoporosis, the benefits of treatment outweigh the risks. Because these agents accumulate in bone with some persistent antifracture efficacy after therapy is stopped, it is reasonable to consider a ‘drug holiday.’ There is considerable controversy regarding the optimal duration of therapy and the length of the holiday, both of which should be based on individual assessments of risk and benefit.

Introduction

Bisphosphonates are widely prescribed for the treatment of osteoporosis. These popular and effective agents have a high affinity for bone and reduce bone resorption by causing loss of osteoclastic resorptive function as well as accelerating osteoclast apoptosis by inhibiting farnesyl pyrophosphate synthase, an enzyme in the HMG-CoA reductase pathway. The four nitrogen-containing bisphosphonates currently in clinical use for the treatment of osteoporosis differ in the strength for binding to bone. The rank order for binding affinity is zoledronate > alendronate > ibandronate > risedronate [Russell et al. 2008]. Higher-affinity bisphosphonates will bind avidly to the bone surface but will spread through bone more slowly, while lower-affinity agents will be distributed more widely through the bone but have a shorter residence time in bone if treatment is stopped. These agents also differ in the potency for inhibiting farnesyl pyrophosphate synthase. The rank order of potency for inhibiting this enzyme is zoledronate > risedronate >> ibandronate > alendronate [Russell et al. 2008]. Bisphosphonate use results in a rapid and substantial decrease in bone turnover markers that is dose and compound dependent, with a maximum effect in 3–6 months. This effect is maintained in a new steady state for at least 10 years with continued treatment [Bone et al. 2004].

Therefore, each bisphosphonate has a unique profile of binding affinity and antiresorptive potency that likely result in clinically meaningful differences in the degree of reduction of bone turnover and the speed of onset and offset of effect.

The skeletal binding sites for bisphosphonates are virtually unsaturable, so a substantial amount could be accumulated over time, leading to a reservoir that continues to be released for months or years after treatment is stopped [Papapoulos and Cremers, 2007]. Since release depends in part on the level of bone turnover, which is reduced by the presence of bisphosphonates, the actual amount released may be fairly small. For example, the amount of alendronate released from bone over the next several months or years after a 10-year treatment period with alendronate would be equivalent to taking one quarter of the usual dose [Rodan et al. 2004]. When treatment is stopped, if there is continued presence of bisphosphonate in bone and continued release (and possible re-attachment to bone), there might be some lingering antifracture effect after treatment is stopped. This is why it is reasonable to consider a ‘drug holiday’ from bisphosphonate therapy, a period of time when treatment is stopped after continuous treatment. However, the term ‘holiday’ implies that treatment will be restarted after some time off.

Studies of long-term use of bisphosphonates

Approval of bisphosphonates in the US was based on studies of 3–4 years duration, although some of these studies have been extended, with alendronate, risedronate and zoledronic acid suggesting efficacy for up to 10 years [Black et al. 2006; Schwartz et al. 2010], 7 years [Mellstrom et al. 2004], and 6 years [Black et al. 2012b], respectively.

The extension of the alendronate Fracture Intervention Trial (FLEX) enrolled subjects who had approximately 5 years of alendronate treatment in the FLEX study into a second 5-year study where subjects were randomized to either continue alendronate or start placebo. At the end of the FLEX study, spine bone mineral density (BMD) increased more (+3.8%) in the long-term treated groups as opposed to the placebo group. Discontinuation of alendronate was associated with gradual increases of bone turnover markers, although at the end of 5 years after stopping alendronate, levels remained somewhat below pretreatment levels 10 years earlier. There were fewer clinical vertebral fractures in the long-term treated group (55% reduction) that met statistical difference from the placebo group (~2% versus 5%, relative risk (RR) 0.45, p = 0.013) [Black et al. 2006]. A post hoc analysis of the FLEX data indicated that nonvertebral fracture risk reduction was also observed in the subset of patients without prior vertebral compression fractures but only in those with T-scores entering FLEX of –2.5 or lower at the femoral neck hip [Schwartz et al. 2010].

The extension of the risedronate VERT-NA study was a 1-year follow up of subjects who completed 3 years of blinded therapy with risedronate or placebo, then stopped their study medications. In the year off treatment, BMD decreased in the former risedronate users, but remained higher than baseline and higher than in the former placebo subjects. Furthermore, bone turnover markers increased and were no different from the former placebo subjects. Despite the apparent resolution of treatment effect on these markers, the risk of new vertebral fractures was reduced by 46% in the former risedronate users compared with the former placebo subjects [Watts et al. 2008]. Similarly, a recent study looking at the effect of discontinuing risedronate for 1 year after 2 or 7 years of treatment also showed decreasing BMD in the total hip and trochanter regions as well as increasing bone turnover markers [Eastell et al. 2011].

In the 3-year extension of the zoledronate HORIZON pivotal fracture trial, subjects who received three doses of zoledronate in the placebo-controlled were assigned at random to one of two arms: a continuation group that received 6 years of zoledronic acid administration and a discontinuation group that received the initial 3 years of zoledronic acid then went to placebo [Black et al. 2012b]. There were small differences in bone density and bone turnover markers in those who continued versus those who stopped treatment, suggesting residual effects. However, there were significantly fewer morphometric vertebral fractures in the group that continued treatment compared with the placebo group (14 versus 30, odds ratio [OR] 0.51, 95% confidence interval [CI] 0.26–0.95, p = 0.035), suggesting that patients at high fracture risk may benefit from continued treatment. Since the majority of subjects in the HORIZON registration trial had prevalent vertebral fractures, the continuation efficacy is most likely confined to those higher-risk patients with vertebral compression fractures [Black et al. 2012b].

In terms of long-term safety of bisphosphonate use, concerns about two uncommon but possible time-related adverse events have emerged: osteonecrosis of the jaw (ONJ) and atypical femur fractures (AFFs). Although no specific issues were identified in the above-mentioned studies, recent evidence supports an association between prolonged bisphosphonate exposure and these two serious conditions. In a retrospective review of the HORIZON trial with intravenous (IV) zoledronate for osteoporosis [Grbic et al. 2010], one case of ONJ was reported in the treatment group and another in the placebo group. A total of 12 fractures in 10 women were classified as subtrochanteric in secondary analyses of the FIT, FLEX and HORIZON Pivotal Fracture Trials, indicating that the risk of such fractures with use of bisphosphonates was very low, even in women who received bisphosphonates for up to 10 years [Black et al. 2010]. Furthermore, iliac crest biopsies after up to 10 years of treatment have not shown oversuppression of bone turnover. Importantly, no causal relationship has been established between prolonged bisphosphonate exposure and either of these outcomes. Even though the risks of ONJ and AFF may increase after 5 years of bisphosphonate therapy, the likelihood remains low. Nevertheless, these safety concerns have led to considerable debate about how long to treat with bisphosphonates.

Drug holidays

On 9 September 2011, the US Food and Drug Administration (FDA) held a hearing to review the long-term safety and efficacy of bisphosphonates including alendronate, risedronate, ibandronate and zoledronate. The majority of the advisory committee (17 to 6) voted that labeling for these drugs should further clarify the duration of use for bisphosphonates but there was a lack of panel consensus on label changes. The FDA wrote their opinion on this perspective, suggesting reevaluation of the need for continuing bisphosphonate therapy beyond 3–5 years in individual patients [Whitaker et al. 2012]. The FDA suggested that a drug holiday may not be advisable in high-risk patients, but for patients discontinuing treatment, there were no concrete recommendations on what should be done.

Subsequently, there has been considerable discussion about who benefits from a drug holiday, when to initiate it, and the ideal duration of the holiday. It is important to note that data from the clinical trials discussed above suggest that the risk of vertebral fractures is reduced beyond 5 years of therapy. In FLEX, the number needed to treat (NNT) for 5 years to prevent one clinical vertebral fracture was 17 in women with a prevalent vertebral fracture and a femoral neck T-score of –2.0 or below at the start of the extension trial, and 24 for women without vertebral fracture and a femoral neck T-score of –2.5 or below [Black et al. 2012a]. Hence, there is evidence for benefit with continued therapy through 10 years in this subset of patients. The 10-year fracture risk assessments with the fracture risk assessment system (FRAX) are increasingly used to guide treatment decisions. A recent study by Leslie and colleagues suggested that the FRAX tool can be used to predict fracture probability in women currently or previously treated for osteoporosis, which may help in guiding the need for continued treatment or treatment withdrawal [Leslie et al. 2012].

Therefore, although there is some residual benefit in terms of fracture reduction for some time after a 3- to 5-year course of bisphosphonate therapy, continuing treatment for 10 years seems to be a better choice for high-risk patients. Even though the risks of bisphosphonate therapy for osteoporosis are small, the risk/benefit ratio may be harmful for low-risk patients. For patients who were candidates for treatment, treatment may be stopped for a drug holiday after a course of some years.

At present, it is difficult to find evidence to support the need for a drug holiday or to establish the effectiveness of treatment after restarting therapy. Similarly, there is no strong evidence to provide guidance in terms of how long to treat, how long the holiday should be, and when the holiday should be stopped. Nevertheless, we believe there is logic to support the following clinical scenarios [Diab and Watts, 2012; Watts and Diab, 2010].

1. Low risk of fracture: treatment is not needed. If a bisphosphonate has been prescribed, it should be discontinued and not restarted unless/until the patient meets treatment guidelines. Example: 54-year-old woman, menopause at age 51, lowest T-score –1.5, no risk factors, bisphosphonate therapy for 3 years. Treatment was not indicated in the first place and can be discontinued.

2. Mild risk of fracture: treat with bisphosphonate for 3–5 years, then stop. The ‘drug holiday’ can be continued until there is significant loss of BMD (i.e. more than the least significant change as determined by the testing center) or the patient has a fracture, whichever comes first. Example: 68-year-old woman, menopause at age 50, initial lowest T-score –2.3, parent with a hip fracture, bisphosphonate treatment for 5 years, BMD stable over that time. Treatment was indicated, but a drug holiday might be considered after 5 years of treatment.

3. Moderate risk of fracture: treat with bisphosphonate for 5–10 years, offer a ‘drug holiday’ of 3–5 years or until there is significant loss of BMD or the patient has a fracture, whichever comes first. Example: 72-year-old woman, menopause at age 48, lowest initial T-score –2.8, no risk factors, bisphosphonate therapy for 7 years, BMD increased over that time so lowest T-score now is –2.3. Treatment was indicated but after 7 years of treatment, a drug holiday might be considered.

4. High risk of fracture (fractures, corticosteroid therapy, very low BMD): treat with bisphosphonate for 10 years, offer a ‘drug holiday’ of 1–2 years, until there is significant loss of BMD or the patient has a fracture, whichever comes first. A nonbisphosphonate treatment (e.g. raloxifene or teriparatide) may be offered during the ‘holiday’ from the bisphosphonate. Example: 75-year-old woman, menopause at age 45, lowest initial T-score –3.6, rheumatoid arthritis requiring ongoing corticosteroid therapy for 12 years, two vertebral fractures by vertebral fracture assessment (VFA), treatment with bisphosphonate therapy for 10 years. Treatment was indicated and she remains at high risk of fracture after 10 years. If a holiday from the bisphosphonate is considered, interval treatment with teriparatide or raloxifene would be prudent.

If a drug holiday is advised, reassessment of risk should occur sooner for drugs with lower skeletal affinity, with a suggestion to reassess after 1 year for risedronate, 1–2 years for alendronate, and 2–3 years for zoledronic acid [Compston and Bilezikian, 2012]. Although it has been proposed that a decrease in BMD or an increase in bone turnover marker (BTM) might be used to decide when to end a drug holiday, there is lack of data on risk for fracture when these surrogate markers begin to change off bisphosphonates. The risedronate study showed that fracture risk remained reduced despite what appeared to be unfavorable changes in these parameters. Conversely, there is no evidence that fracture risk is reduced if BMD is stable or BTM is low off treatment. That being said, in clinical practice, monitoring BMD and BTM are the only means of gaining some sense of the loss of the effect of the bisphosphonate on bone remodeling, but ultimately the duration of the holiday should be based on clinical judgment.

Conclusion

In conclusion, bisphosphonates that have been approved for the treatment of postmenopausal osteoporosis are effective and generally safe agents that have robust evidence for fracture risk reduction. Their systemic safety is related to their binding to bone and lack of uptake by other tissues other than the kidney. A reservoir of bisphosphonates accumulates after years of treatment that is gradually released over months or years and appears to result in a lingering antifracture benefit for some time after therapy is stopped. This makes it possible to consider ‘drug holidays’, time off bisphosphonate therapy (but possibly on another agent), and then resuming therapy. Although there is no strong evidence to guide us, we believe that some time off treatment should be offered to most patients on long-term bisphosphonate therapy. The duration of treatment and the length of the ‘holiday’ should be tailored to individual patient circumstances and based on individual assessments of risk and benefit. In the higher-risk population treated for the right duration, bisphosphonates have an exceptionally high benefit/risk ratio. While lower-risk patients may be offered a ‘drug holiday’ after 3–5 years of use, higher-risk patients should be counseled on the greater risk for fracture if discontinuation is initiated. The strength of the evidence for fracture reduction in high-risk individuals and the rarity of long-term adverse effects indicate that the benefits of continued treatment outweigh the risks in individuals at high risk of fracture.