Discontinuation of odanacatib and other osteoporosis treatments: Here today and gone tomorrow?

The options available for effective osteoporosis treatment continue to grow.1 Because of the large number of treatment options, many clinicians, particularly in primary care, are uncertain about when to start therapy, which drug to use, and when to stop. The issue of when to stop antifracture therapy is now front and center as evidence accumulates that the antifracture benefits of some treatments persist after discontinuation2 and as patients increasingly voice concerns about possible rare but serious problems linked to osteoporosis treatment, such as atypical femoral fractures, osteonecrosis of the jaw, arrhythmias, and cancer.3 The decision about when to stop or change osteoporosis treatment is particularly complex and should be influenced by the expected benefits and harms of continued use and at least some information about what to expect after therapy is stopped (also known as resolution-of-effect). A prolonged skeletal resolution-of-effect, characteristic of at least some bisphosphonates, might be desirable if ongoing antifracture benefit is desired after discontinuation. Conversely, a rapid resolution of effect, which is typical of sex hormone and selective estrogen receptor modulator (SERM) therapy, might be preferable if there are concerns about side effects with prolonged exposure or blunting of the benefits from subsequent treatment with other agents. Rapid resolution-of-effect also might be advantageous for women of child-bearing age who plan to conceive after treatment for osteoporosis and for prophylaxis against glucocorticoid-induced bone loss when the duration of steroid treatment is finite. If for any reason discontinuation of treatment transiently increases fracture risk, even short-term interruptions (or intermittent noncompliance) may have adverse clinical consequences. The preferred approach to generate resolution-of-effect data is to extend recently completed placebo-controlled treatment trials and rerandomize individuals from the active treatment arm to either continue active therapy or be switched to matching placebo. Ideally, the original placebo group also should continue to receive placebo, but for efficacious treatments this is usually neither ethical nor feasible. This approach provides clinically useful information about the effects of discontinuation, particularly when the studies are blinded and sufficiently large to assess fracture outcomes. Because fracture endpoint resolution-of-effect investigations are difficult and expensive, few osteoporosis treatments have been evaluated in this way. To date, the Long-Term Extension of the Fracture Intervention Trial (FLEX)4 is the only published randomized, placebo-controlled trial sufficiently large to report fracture outcomes among those who continue versus those who discontinue osteoporosis treatment. In FLEX, the risk of clinical spine fracture was 50% lower among women who continued alendronate, but the risk of morphometric spine and nonspine fracture was similar among women who continued alendronate for up to 10 years compared with those who discontinued alendronate after 5 years. It is important to remember that the FLEX study did not include women who received placebo for all 10 years, and the similar rates of fracture in the continuation and discontinuation groups does not exclude the possibility of an increase in fracture risk in both groups. A recent study of discontinuation of zolendronic acid reported fracture results that were similar to FLEX, but those data are not yet published.5 A much less expensive but also much less useful alternative to resolution-of-effect studies with fracture endpoints is to rely on surrogate endpoints, such as changes in bone mineral density (BMD) and/or biochemical markers of bone turnover. Unfortunately, little is known about how short-term changes in these surrogate markers after discontinuation of osteoporosis treatment relate to long-term effects on fracture outcomes. The results of a surrogate endpoint resolution-of-effect study are published in this issue of JBMR.6 Odanacatib inhibits cathepsin K, a cysteine protease produced by osteoclasts that degrades type 1 collagen. Since cathepsin K is relatively specific to osteoclasts, it represents a promising candidate for drug development, and preclinical studies have demonstrated that odanacatib has profound effects on bone resorption with few off-target effects.7 Odanacatib is currently being examined in a large trial with fracture endpoints, with the eventual goal of Food and Drug Administration (FDA) approval for use in individuals with osteoporosis, but smaller trials have found that odanacatib treatment is associated with significant reductions in bone turnover and increases in axial BMD.8, 9 In the current odanacatib study,6 Eisman and colleagues describe the results of an extension of a published phase II (dose-ranging) trial. In the parent study,9 399 postmenopausal women with low bone mass were randomized to placebo or one of four doses of weekly odanacatib and followed for 2 years. As expected from previous studies, bone turnover fell and BMD increased in a dose-dependent fashion. In the extension study, 189 women who completed the parent study were rerandomized without unblinding to an additional year of either 50 mg/week of odanacatib or matching placebo. Since 50 mg/week is being evaluated in the phase III odanacatib studies, Eisman and colleagues focus on two groups: those who received odanacatib 50 mg/week for all 3 years and those who received 2 years of odanacatib followed by 1 year of placebo. In the group receiving 3 years of odanacatib, axial BMD continued to increase, and bone turnover remained relatively stable during the last year, whereas discontinuation of odanacatib after 2 years resulted in dramatic increases in bone turnover and rapid bone loss. For example, in the year following discontinuation of odanacatib, serum cross-linked N-telopeptide of type I collagen (NTX) increased 230%, and spine BMD fell about 6%. The odanacatib resolution-of-effect study reported by Eisman and colleagues6 is well designed, and the results are interesting, but several important limitations are worth noting. The most important shortcomings of the study are the small sample size and short duration. Among the participants in the parent study who received 2 years of odanacatib 50 mg/week, only 20 women were rerandomized to an additional year of odanacatib, and only 18 women received an additional year of placebo. Such small numbers and short follow-up allow detection of only large changes in bone turnover and BMD and are inadequate to address fracture risk. In addition to the usual extension study caveats about the proportion and characteristics of the participants who elect to continue, it is unfortunate that only a per-protocol analysis is reported (and not intention-to-treat) and that the industry sponsor was unblinded to treatment allocation after the first year of follow-up. Lastly, some of the resorption marker data in the study may be difficult to interpret because the collagen degradation products that result from cathepsin K are known to differ from those produced by the activity of other osteoclast enzymes.10 Although fracture outcomes are preferred, it is of some interest to compare the odanacatib resolution-of-effect data with results from studies of other potential first-line osteoporosis treatments. These comparisons should be interpreted with caution: The studies are not head-to-head examinations, and therefore the populations studied, severity of disease, and analytical approach may differ substantially. Nonetheless, the available evidence suggests that standard doses of alendronate, risedronate, denosumab, and odanacatib have similar effects on bone turnover and BMD after 2 years of treatment, but discontinuation of treatment results in substantially greater 1-year axial bone loss among denosumab- and odanacatib-treated women compared with the bisphosphonate-treated women (Table 1). Discontinuation of odanacatib in particular appears to increase bone markers substantially, a potentially worrisome scenario if one accepts the current paradigm that rapid bone loss and sustained increases in bone turnover are independently associated with increased fracture risk, at least among older untreated women.11, 12 Whether these apparently different short-term effects on bone turnover and bone loss translate into meaningful effects on fracture risk can be determined only with much larger and longer trials. = 437)13 = 398)14 = 128)15 = 20)6 In sum, resolution-of-effect data are critical for the optimal use of effective osteoporosis treatments, and such data are beginning to appear for many potential first-line treatments. Such data are best obtained from carefully executed and blinded studies of sufficient size and duration to address effects on both efficacy and potential side effects. Although most published resolution-of-effect studies address only surrogate outcomes, such as bone loss and changes in bone turnover, our knowledge of how these surrogates relate to fracture outcomes when therapy is stopped is incomplete. Extrapolating from untreated individuals, rapid bone loss and dramatic increases in bone turnover after discontinuation could increase fracture risk transiently. The risk of fracture after discontinuation of osteoporosis medications with rapid resolution of effect is particularly important because intended and unintended interruptions in therapy, including noncompliance, are very common in clinical practice. The author states that he has consulted for Merck and received research support from Novartis and Amgen.