This report is available in PDF only (Full Report [952 KB] and Appendixes [1.8 MB]). People using assistive technology may not be able to fully access information in these files. For additional assistance, please contact us.
Purpose of Review
To assess the current evidence about the efficacy and harms of long-term osteoporosis drug therapy (>3 years) to prevent fractures, of osteoporosis drug continuation versus discontinuation (placebo drug holiday), and whether osteoporosis drug effects differ by patient, bone, and drug characteristics.
- Evidence about efficacy and harms of long-term osteoporosis drug therapy is predominately from U.S.-based studies conducted in white, postmenopausal women.
- In postmenopausal women with osteoporosis, 4 years of alendronate compared with placebo reduces risk of incident clinical fractures and incident radiographic vertebral fractures; whereas, among women with osteopenia, it does not reduce risk of incident clinical fractures and may not reduce risk of incident radiographic vertebral fractures.
- In postmenopausal women with osteoporosis, 4 years of raloxifene compared with placebo reduces risk of incident radiographic vertebral fractures and incident clinical vertebral fractures, but does not lower risk of incident hip or other nonvertebral fractures.
- In postmenopausal women with a history of hysterectomy and past clinical fractures, but unknown bone mineral density (BMD), 7 years of estrogen versus placebo reduces risk of incident clinical fractures and incident hip fractures.
- In postmenopausal women with an intact uterus and past clinical fractures, but unknown bone mineral density (BMD), 5.6 years of estrogen/progestin versus placebo reduces risk of incident clinical fractures.
- Continued use of zoledronic acid for a total of 6 years versus discontinuation after 3 years lowers risk of incident radiographic vertebral fractures, but does not lower risk of incident nonvertebral fractures.
- Continued use of alendronate for a total of 10 years versus discontinuation after 5 years probably lowers risk of incident clinical vertebral fractures, but does not lower risk of incident nonvertebral fractures.
- Long-term bisphosphonate use may increase risk of atypical femoral fractures with confirmed radiological features (AFF), subtrochanteric or femoral shaft fractures without confirmed radiological AFF features, and osteonecrosis of the jaw (ONJ).
- Future trials should include diverse populations (including male sex, nonwhite race, and individuals with multiple comorbidities), include individuals with high fracture risk who do not meet criteria for osteoporosis, be large enough to compare between-treatment differences in risk of incident clinical fractures, and compare sequential treatments (e.g., anabolic followed by anti-resorptive) and drug holidays of various durations.
- Future observational studies on risk of AFF and ONJ should use a cohort design and define outcomes using standard radiographic criteria.
Objective. To summarize evidence on outcomes of long-term osteoporosis drug therapy to prevent fractures, on continuing versus discontinuing osteoporosis drug therapy (i.e., placebo drug holidays), and on whether osteoporosis drug intervention effects vary as a function of patient, bone, or drug characteristics.
Data sources. MEDLINE, Embase and Cochrane databases from 1995 to June 2018; ClinicalTrials.gov; bibliographies of relevant systematic reviews
Review methods. Long-term osteoporosis drug therapy was defined as >3 years and drug holiday as osteoporosis drug discontinuation for ≥1 year after ≥1 year of prior osteoporosis drug use. Two reviewers independently rated risk of bias (ROB) and strength of evidence (SOE), resolving discrepancies by consensus. Included studies were English-language trials for incident fractures and harms and controlled observational studies for additional harms outcomes. For low or medium ROB studies, one reviewer extracted data and a second verified accuracy.
Results. Of 56 eligible publications, 44 had low or medium ROB, including 32 publications of trials (7 unique studies) and 12 publications of observational studies (10 unique studies). Nearly all studies were comprised of postmenopausal women. Mean participant age was 72 years; all but 2 studies had a mean age <80 years. In postmenopausal women with osteoporosis, compared with placebo, 4 years of alendronate or raloxifene reduced risk of incident vertebral fractures (high SOE), 4 years of alendronate reduced risk of incident clinical fractures (moderate SOE). In postmenopausal women with past fractures, compared with placebo, both long-term estrogen and long-term estrogen/progestin reduced risk of incident clinical fractures and long-term estrogen reduced risk of incident hip fractures (all low SOE). Alendronate, denosumab and raloxifene for 4 years each significantly increased total hip and lumbar spine bone mineral density (BMD) versus placebo. Continuation versus discontinuation of alendronate after 5 years reduced risk of incident clinical vertebral fractures in one large trial (10 vs. 5 years, moderate SOE), but not in another smaller trial (7 vs. 5 years, low SOE). Continuation versus discontinuation of zoledronic acid (6 vs. 3 years) reduced risk of incident radiographic vertebral fractures (moderate SOE), but evidence was insufficient about risk of incident clinical vertebral fractures. Neither alendronate nor zoledronic acid continuation reduced risk of nonvertebral fractures (low SOE) versus discontinuation; for both, continuation was associated with generally stable hip BMD compared to small, but significant declines with discontinuation. Based primarily on observational studies, long-term bisphosphonates may increase risks of radiologically confirmed atypical femoral fractures (AFF), subtrochanteric or femoral shaft fractures without confirmed AFF features, and osteonecrosis of the jaw (ONJ).
Limitations. Minimal data for men or individuals with comorbidities. Low power to assess risks of incident clinical fractures. No data compared long-term effects of sequential treatments (e.g., anabolic followed by anti-resorptive) or different durations of drug holidays. Analyses of possible treatment effect modifiers almost entirely post hoc. Observational studies used variable drug treatment and control exposures and harms definitions.
Conclusions. For postmenopausal women with osteoporosis by BMD or past fractures, long-term alendronate and raloxifene reduced risk of incident vertebral fractures; long-term alendronate, estrogen, and estrogen/progestin reduced risk of clinical fractures; and long-term estrogen reduces risk of incident hip fractures. Longer-term use of bisphosphonates versus discontinuation may lower vertebral fracture risk and stabilize hip BMD, but doesn’t reduce nonvertebral fracture risk and may increase risk of AFF and ONJ. Long-term estrogen and estrogen/progestin increased risk of cardiovascular disease, and long-term estrogen increased risk of dementia and breast cancer. To address remaining knowledge gaps, future trials and observational studies should enroll diverse populations (sex, comorbidity), examine the effects of sequential treatments and compare drug holidays of different durations, be powered for clinical fractures, and use standard AFF and ONJ definitions. A priori analyses to examine whether treatment outcomes vary by patient, bone and drug treatment characteristics may inform individualized treatment decisions.
Suggested citation: To be provided in the Final Report.