Long-Term Drug Therapy and Drug Discontinuations and Holidays for Osteoporosis Fracture Prevention

Abstract

Osteoporosis is a skeletal disorder of low bone mass and microarchitectural deterioration of bone that leads to bone fragility and increased risk for fracture (1). About 10 million U.S. adults aged 50 years or older have osteoporosis (2), and about 2 million U.S. adults experience an osteoporotic or other low-trauma fracture each year (3). Such fractures often cause pain, disability, and impaired quality of life (4, 5), and hip fractures and clinical vertebral fractures are associated with increased mortality (5, 6). Most fracture risks increase sharply with age; therefore, fracture burden is projected to increase in coming decades as the population ages. Several osteoporosis drug treatments (ODT) reduce fractures in short-term randomized controlled trials of up to 3 years. Bisphosphonates, denosumab, teriparatide, and abaloparatide reduce nonvertebral fractures and clinical and radiographic vertebral fractures (7, 8). Bisphosphonates and denosumab also lower risk for hip fractures (8). Less is understood about the benefits and harms of initiating long-term ODT or, in patients who have already completed short-term treatment, of continuing versus discontinuing ODT. A recent American College of Physicians guideline recommended ODT with a bisphosphonate or denosumab for 5 years to reduce hip and vertebral fractures in osteoporotic women but suggested that high-risk patients may benefit from longer treatment (8). Concerns that long-term bisphosphonate use might increase fracture risk by inhibiting normal repair of bone microdamage (9, 10) have led to suggestions to stop bisphosphonate treatment and to restart it or another ODT later (called an ODT drug holiday) (9). Several groups advocate bisphosphonate drug holidays to minimize harms while preserving as much antifracture benefit as possible. However, consensus is lacking around which patients should have bisphosphonate holidays, when, and for how long, as well as criteria for restarting treatment (9, 11, 12). In contrast, drug holidays are not recommended for denosumab users because bone loss increases rapidly after discontinuation, which may increase risk for vertebral fractures (13). To focus on these uncertainties, this systematic review addressed 4 questions. First, what are the effects of long-term (>3 years) ODT versus control on risks for incident fractures and harms? Second, do the effects of long-term ODT vary as a function of patient, bone, or drug characteristics? Third, among patients receiving ODT to prevent fracture, what are the effects of continuing versus at least temporarily stopping treatment on risks for incident fractures and harms? Fourth, do outcomes of ODT continuation versus discontinuation vary as a function of patient, bone, or drug characteristics? Methods We developed and followed a standard protocol, which is registered in PROSPERO (CRD42018087006) and available at https://effectivehealthcare.ahrq.gov/topics/osteoporosis-fracture-prevention/research-protocol. The full technical report contains search strategies, flow diagrams, evidence tables, study quality assessment tables, and detailed results (https://effectivehealthcare.ahrq.gov/topics/osteoporosis-fracture-prevention/research). Data Sources and Searches We searched MEDLINE, Embase, and the Cochrane Library for studies published from January 1995 through October 2018. We searched bibliographies of relevant systematic reviews published since 2012 and articles suggested by experts. We searched ClinicalTrials.gov for additional studies. Study Selection We included English-language studies that enrolled postmenopausal women or men aged 50 years or older who were being investigated or treated for fracture prevention. Studies had to evaluate ODTs approved by the U.S. Food and Drug Administration and compare long-term ODT (>3 years) versus control or ODT continuation versus discontinuation (cessation for 1 year after 1 year of use). Trials reporting incident fractures must have reported results for participants with osteoporosis or osteopenia, whereas controlled observational studies reporting rare harms (such as atypical femoral fracture [AFF]) were not restricted by osteoporosis or osteopenia status. Populations with known secondary causes of osteoporosis were excluded. Two reviewers independently examined articles for eligibility and resolved discrepancies by consensus. Data Extraction and Quality Assessment For each eligible study, 2 reviewers independently rated risk of bias (ROB) for outcomes of interest as low, medium, or high on the basis of criteria from the Agency for Healthcare Research and Quality. For articles with low or medium ROB, 1 reviewer extracted details on study design, inclusion criteria, participant characteristics, interventions, treatment duration, and incident fractures and harms, and a second reviewer verified accuracy. Two reviewers graded strength of evidence (SOE) as high, moderate, low, or insufficient on the basis of study limitations, directness, consistency, and precision. We confirmed ROB and SOE assessments by consensus. Data Synthesis and Analysis We synthesized data qualitatively because studies had heterogeneous patient populations, intervention (exposure) and control groups, and definitions of incident fractures and harms. We considered studies that compared long-term ODT versus continuous (inactive) control separately from those comparing long-term ODT versus shorter-term ODT followed by discontinuation. Role of the Funding Source This review was funded by the National Institutes of Health (NIH) Office of Disease Prevention through an interagency agreement with the Agency for Healthcare Research and Quality. Agency staff, an NIH Office of Disease Prevention working group, an NIH content area expert group, and a technical expert panel helped refine the project scope. The draft report was presented at an NIH Office of Disease Prevention Pathways to Prevention workshop. Results We identified 8356 unique publications through October 2018, of which 61 met eligibility criteria and were included in this review. Of the 48 eligible publications with low or medium ROB, 35 were trials (9 unique studies, including 7 of long-term ODT vs. placebo and 2 of ODT continuation vs. discontinuation without resumption) and 13 were controlled observational studies (11 unique studies, including 10 of long-term ODT vs. control and 1 of ODT continuation vs. discontinuation without resumption). Only 1 study included a treatment group of ODT followed by discontinuation and resumption (holiday), but results for its ODT holiday group were pooled with those of its continuous treatment and discontinuation groups. All trials enrolled only postmenopausal women, most with osteoporosis defined by bone mineral density (BMD) or vertebral fracture history and some with osteopenia defined by BMD. Observational studies included 84% to 100% women. No observational studies reported BMD, but many enrolled participants with past fractures or ODT. Mean participant age was 72 years. Long-Term ODT Efficacy In women with osteopenia or osteoporosis defined by BMD without existing vertebral fracture, 4 years of alendronate compared with placebo reduced radiographic vertebral fractures (hazard ratio [HR], 0.56 [95% CI, 0.39 to 0.80]) (high SOE) but did not significantly reduce nonvertebral fractures (HR, 0.88 [CI, 0.74 to 1.04]) (low SOE) or hip fractures (HR, 0.79 [CI, 0.43 to 1.44]) (low SOE) (14). In women with osteoporosis defined by BMD or past vertebral fracture, 4 years of raloxifene compared with placebo reduced radiographic vertebral fractures (relative risk, 0.64 [CI, 0.53 to 0.76]) and clinical vertebral fractures (relative risk, 0.58 [CI, 0.43 to 0.79]) (both high SOE) (15) but not nonvertebral or hip fractures (high or moderate SOE, respectively). In older women with osteopenia or osteoporosis, 6 years of zoledronic acid compared with placebo reduced clinical fractures (HR, 0.73 [CI, 0.60 to 0.90]) (moderate SOE), including nonvertebral fractures (HR, 0.66 [CI, 0.51 to 0.85]) (high SOE) and clinical vertebral fractures (HR, 0.41 [CI, 0.22 to 0.75]) (moderate SOE) (16). In women with unspecified osteoporosis or osteopenia status, compared with placebo, both estrogenprogestin for 5.6 years (17) and unopposed estrogen for 7 years (18) reduced risk for clinical fractures (high SOE), including hip fractures (moderate SOE) (Table 1) (4, 1432). Table 1. Efficacy of Long-Term Osteoporosis Drug Treatment (>3 Years) Evidence was insufficient to compare fracture risk between women receiving long-term denosumab therapy and those receiving placebo (19), and we identified no eligible studies addressing the long-term antifracture efficacy of sequential ODTs (for example, anabolic followed by antiresorptive therapy or denosumab followed by bisphosphonate). Variation in Efficacy by Patient, Bone, or Drug Characteristics Antifracture efficacy of long-term alendronate therapy varied as a function of baseline BMD (14). In 1 randomized controlled trial, 4 years of alendronate reduced clinical fractures (HR, 0.64 [CI, 0.50 to 0.82]) and radiographic vertebral fractures (HR, 0.50 [CI, 0.31 to 0.82]) in women with osteoporosis defined by BMD only, but not in women with osteopenia (P for interaction= 0.01 for clinical fracture; not reported for radiographic vertebral fracture). In a post hoc analysis, 4 years of alendronate also reduced hip fractures in women with osteoporosis defined by BMD only but not in women with osteopenia (14). In additional post hoc analysessome in osteopenic subgroups (33, 34)past nonvertebral fracture (34), 10-year probability of major osteoporotic fracture (35), and pretreatment levels of bone turnover markers (36) did not modify the long-term effect of alendronate compared with placebo on risk for any incident fracture. Whether the antifracture efficacy of zoledronic acid varies as a function of baseline BMD is unclear. In secondary analyses from a 6-year trial in older women