To review evidence about screening to prevent osteoporotic fractures for the U.S. Preventive Services Task Force (USPSTF).
PubMed, the Cochrane Library, Embase, and trial registries from November 1, 2009, through October 1, 2016, and surveillance of the literature through March 23, 2018; bibliographies from retrieved articles.
Two investigators independently selected studies using a priori inclusion and exclusion criteria. We selected studies with a majority of adults age 40 years or older conducted in countries with a very high human development index. For screening studies, we required that studies include a majority of participants without prevalent low-trauma fractures. For treatment studies, we required that studies include a majority of participants with increased fracture risk. We selected studies of screening tests (fracture risk prediction instruments, bone measurement testing, or a combination of fracture risk prediction instruments and bone measurement testing) that were feasible for primary care settings and available in the United States. We selected studies of treatment approved by the U.S. Food and Drug Administration for synthesis of benefits and harms. We excluded studies of poor quality and of fracture risk prediction instruments without external validation.
One fair-quality trial demonstrated reduction in hip fractures when comparing screening with no screening (2.6% v 3.5%, Hazard rate [HR] 0.72; 95% confidence interval [CI], 0.59 to 0.89). The study reported no other statistically significant benefits (osteoporotic or clinical fractures, mortality) or harms (anxiety, quality of life). We included 168 articles of fair or good quality; 105 articles assessed screening accuracy and 65 articles assessed benefits and harms of treatment. Using centrally measured dual-energy X-ray absorptiometry (DXA) as the reference standard for identifying osteoporosis, the pooled estimate of accuracy as measured by the area under the curve (AUC) for clinical risk assessment instruments for women ranges from 0.65 to 0.76 and for men from 0.76 to 0.80. AUCs for the accuracy of calcaneal quantitative ultrasound in identifying central DXA—measured osteoporosis for women is 0.77 (95% CI, 0.72 to 0.82, 7 studies) and for men is 0.80 (95% CI, 0.67 to 0.94, 3 studies). The AUCs of machine-based tests, including centrally measured DXA (areal bone mineral density and trabecular bone score) and calcaneal quantitative ultrasound, for predicting fractures ranged from 0.59 to 0.86 (21 studies). The AUCs for instruments predicting fractures, some of which incorporate machine-based tests, have similar accuracy (pooled AUC range for the Fracture Risk Assessment Tool: 0.62 to 0.79; 24 studies). Available but limited evidence in studies including participants with a wide spectrum of baseline bone mineral density from normal to osteoporosis suggests no benefit from repeating a bone measurement test between 4 and 8 years after the initial screen. Evidence from placebo-controlled trials demonstrates the following benefits. For women, the risk of vertebral fractures can be reduced by bisphosphonates, parathyroid hormone, raloxifene, and denosumab by 36 percent to 68 percent. Relative risks (RRs) range from 0.32 (parathyroid hormone or denosumab) to 0.64 (raloxifene). The risk of nonvertebral fractures can be reduced by 16 percent to 20 percent by bisphosphonates and denosumab (RR, 0.84 and 0.80, respectively). The risk of hip fractures can be reduced by 40 percent by denosumab (RR, 0.60). Evidence from bisphosphonates does not demonstrate benefit for hip fractures. Evidence is very limited for men. The risk of morphometric vertebral fractures can be reduced by 67 percent by zoledronic acid (RR, 0.33). No studies demonstrate reductions in risk of clinical vertebral fractures or hip fractures for men. Evidence on variations in effectiveness for subgroups is also limited; a single trial each for five drugs suggests no differences in effectiveness by age, baseline bone mineral density, prior fractures, or a combination of risk factors. Bisphosphonates are not consistently associated with discontinuations, serious adverse events, gastrointestinal events, or cardiovascular events. No included studies reported cases of osteonecrosis of the jaw or atypical femur fracture, although evidence from excluded studies (including active comparisons, case series, and secondary prevention populations) suggests an increased but rare risk of these outcomes. Raloxifene increases the risk of deep vein thrombosis (0.7% vs. 0.3%, RR, 2.14; 95% CI, 0.99 to 4.66; I2=0%, 3 studies, N=5,839) and hot flashes (11.2% vs. 7.6%, RR, 1.42; 95% CI, 1.22 to 1.66; I2=0%, 5 trials; N=6,249) when compared with placebo.
The evidence is limited on the direct question of the benefits and harms of screening for elevated osteoporotic fracture risk. The indirect evidence pathway rests on studies evaluating (1) the accuracy of screening approaches in identifying osteoporosis and predicting fractures and (2) the benefits of treatment among those with osteoporosis or at high risk for fractures. Other limitations of the evidence base relate to underlying heterogeneity in baseline risk, prior fractures, prior treatment, and duration of followup.
Evidence from one trial of screening to prevent osteoporotic fractures suggests a reduction in hip fractures. The accuracy of clinical risk assessment tools for identifying osteoporosis or predicting fractures generally ranges from very poor (0.50) to good (0.90). Treatments reduce the risk of vertebral and nonvertebral fractures. Studies do not consistently demonstrate an increased risk of harms for drugs, although studies of raloxifene suggest a trend toward higher risk of deep vein thrombosis. Rare harms, such as osteonecrosis of the jaw and atypical femur fractures were not reported in this body of evidence but they may occur. The evidence is limited for subpopulations characterized by age, sex, baseline bone mineral density, and baseline fracture risk.