Screening for Osteoporosis to Prevent Fractures: A Systematic Evidence Review for the US Preventive Services Task Force

Abstract

Importance:

Fragility fractures result in significant morbidity.

Objective:

To review evidence on osteoporosis screening to inform the US Preventive Services Task Force.

Data Sources:

PubMed, Embase, Cochrane Library, and trial registries through January 9, 2024; references, experts, and literature surveillance through July 31, 2024.

Study Selection:

Randomized clinical trials (RCTs) and systematic reviews (SRs) of screening; pharmacotherapy studies for primary osteoporosis; predictive and diagnostic accuracy studies.

Data Extraction and Synthesis:

Two reviewers assessed titles/abstracts, full-text articles, study quality and extracted data; when at least 2 similar studies were available, meta-analyses were conducted.

Main Outcomes and Measures:

Hip, clinical vertebral, major osteoporotic, and total fractures; mortality; harms; accuracy.

Results:

Three RCTs and 3 SRs reported benefits of screening in older, higher-risk women. Two RCTs used two-stage screening: Fracture Risk Assessment Tool [FRAX^®] estimate with bone mineral density (BMD) testing if risk threshold exceeded. One RCT used BMD plus additional tests. Screening was associated with reduced hip (pooled relative risk [RR], 0.83 [95% confidence interval {CI}, 0.73 to 0.93]; 3 RCTs; 42 009 participants) and major osteoporotic fracture (pooled RR, 0.94 [95% CI, 0.88 to 0.99]; 3 RCTs; 42 009 participants) compared with usual care. Corresponding absolute risk differences were 5 to 6 fewer fractures per 1000 participants screened.

The discriminative accuracy of risk assessment instruments to predict fracture or identify osteoporosis varied by instrument and fracture type; most had an area under the curve between 0.60 and 0.80 to predict major osteoporotic fracture, hip fracture, or both. Calibration outcomes were limited.

Compared with placebo, bisphosphonates (pooled RR, 0.67 [95% CI, 0.45 to 1.00]; 6 RCTs; 12 055 participants) and denosumab (RR, 0.60 [95% CI, 0.37 to 0.97] from largest RCT with 7808 participants) were associated with reduced hip fractures. Compared with placebo, no statistically significant associations were observed for adverse events.

Conclusions and Relevance:

Screening in higher-risk women age 65 or older was associated with a small absolute risk reduction in hip and major fractures compared with usual care. No evidence evaluated screening with BMD alone or screening in men or younger women. Risk assessment instruments, BMD alone, or both have poor to modest discrimination for predicting fracture. Osteoporosis treatment with bisphosphonates or denosumab over several years was associated with fracture reductions and no meaningful increase in adverse events.

INTRODUCTION

The primary rationale for screening for osteoporosis is to identify persons who would benefit from pharmacotherapy to reduce the incidence and morbidity from fragility fractures, which are defined as fractures resulting from low-energy trauma (eg a fall from standing height or less). In 2018, the US Preventive Services Task Force (USPSTF) recommended screening with bone measurement testing in women aged 65 years or older and in postmenopausal women younger than 65 years with increased risk of osteoporosis as determined by a formal clinical risk assessment tool.¹ The evidence was insufficient for the USPSTF to assess the benefits and harms of screening in men (I statement).¹ This update review evaluated the current evidence on screening to inform an updated recommendation by the USPSTF.

METHODS

Scope of the Review

The review was guided by the analytic framework and key questions (KQs) shown in Figure 1. Additional details are provided in the full evidence report (add final URL).

Figure 1. Analytic Framework: Screening for Osteoporosis to Prevent Fractures.

Evidence reviews for the USPSTF use an analytic framework to visually display the key questions that the review will address in order to allow the USPSTF to evaluate the effectiveness and safety of a preventive service. The questions are depicted by linkages that relate interventions and outcomes. For further details see the USPSTF Procedure Manual.¹² DXA indicates dual energy X-ray absorptiometry.

Data Sources and Searches

PubMed, Embase, and the Cochrane Library were searched for studies published in English from April 1, 2016, through January 9, 2024. ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform were also searched. To supplement electronic searches (Appendix B.1 in the Full Report), reference lists of relevant articles and reviewer-suggested studies were reviewed. As part of ongoing surveillance, targeted journal searches were conducted to identify major studies that might affect the conclusions of the evidence and the related USPSTF recommendation. The last surveillance was on July 31, 2024.

Study Selection

Titles, abstracts, and full-text articles were reviewed by 2 independent reviewers using prespecified criteria for each KQ (Appendix B2 in the Full Report); disagreements were resolved by discussion or a third reviewer.

For KQ1 and 3, randomized clinical trials (RCTs) and systematic reviews (SRs), or controlled cohort studies (for harms only) comparing screening by dual x-ray absorptiometry (DXA) testing, risk assessment (eg, Fracture Risk Assessment Tool [FRAX^®] =), or both compared with no screening among persons without known osteoporosis (i.e., T-score BMD < −2.5) were eligible. For KQ4 and 5, RCTs or controlled cohort studies (for harms only) evaluating FDA-approved bisphosphonates or denosumab compared with placebo or no treatment were eligible if most enrolled participants did not have secondary osteoporosis or prior fragility fractures to approximate a screen-detected population. Studies of other drugs in men were also eligible. Outcomes for KQ1, 3, 4, and 5 included fractures, mortality, and harms.

For KQ2, primary studies or SRs reporting discrimination or calibration outcomes were eligible if they 1) evaluated risk assessment instruments (KQ2a) or BMD alone (KQ2b) for predicting future incident fracture or 2) evaluated the accuracy of risk assessments for identifying current osteoporosis (KQ2c). Only risk assessments evaluated in at least 2 independent cohorts external to the development cohort were eligible for KQ2a and 2c, except if conducted in men.

Studies included in the prior 2018 review for the USPSTF² were reassessed against the study selection and methodological quality criteria for this update. Fair- or good-quality studies that met all study selection criteria and were conducted in countries categorized as very highly developed by the 2020 United Nations Human Development Index³ were eligible. However, for KQ2a and 2b, eligibility was further limited to countries with similar hip fracture incidence as the United States (moderate category⁴) but poor-quality studies were included because of the limited pool of good- or fair-quality predictive accuracy studies.

Data Extraction and Quality Assessment

For each included study, 1 reviewer abstracted relevant characteristics (ie, population, intervention, comparator) and outcome data into a structured form and a second reviewer checked data for accuracy.

The risk of bias (ROB) for each included study was assessed by 2 independent reviewers using design-specific ROB tools (ROB 2 for RCTs,⁵ ROBINS-I for nonrandomized studies of interventions,⁶ QUADAS-2 for diagnostic test accuracy,⁷ PROBAST for predictive accuracy,^8–10 and ROBIS for SRs¹¹). For predictive accuracy of studies evaluating risk assessment instruments, the ROB of each instrument in its development cohort(s) was evaluated using the full PROBAST instrument⁸ but adapted to include health equity signaling items. The ROB for instruments evaluated in external validation cohorts was assessed using an adapted version of the PROBAST short form.¹⁰ ROB ratings from design-specific instruments were translated to methodological quality ratings using predefined criteria developed by the USPSTF.¹² Disagreements in study quality ratings were resolved through discussion.

Data Synthesis and Analysis

Data were synthesized in narrative and tabular formats. When at least 2 similar studies were available, a quantitative synthesis was performed using random effects models with the inverse-variance weighted method of DerSimonian and Laird¹³ in Stata version 17 (StataCorp) to generate pooled estimates of the relative risk (RR), which were reexpressed as absolute risk differences (ARDs) per 1000 persons screened or treated.¹⁴ Statistical heterogeneity was assessed by the I² statistic.^15,16 For KQ4 and KQ5, data was pooled across dosage groups for studies with more than 1 active intervention arm. Sensitivity analyses were conducted for alternative types of vertebral fracture (clinical vs radiographic), for non-FDA-approved dosages of drugs, and with alternative methods of pooling for outcomes with rare (< 1%) or 0 events in either study arm.¹⁷

The strength of evidence (SOE) for each comparison and outcome was assessed as high, moderate, low, or insufficient using methods developed for the USPSTF and the Agency for Healthcare Research and Quality Evidence-based Practice Center program.^18,19 Two senior reviewers independently developed SOE assessments; disagreements were resolved through discussion.

RESULTS

One hundred and forty-five unique studies (published in 195 articles) were included in this update (Figure 2). A list of full-text articles that were reviewed but excluded is in Appendix C of the Full Report.

Figure 2. Literature Search Flow Diagram: Screening for Osteoporosis.

HDI indicates human development index; KQ, key question; RCT, randomized clinical trial; SR, systematic review.

Benefits and Harms of Screening

Key Question 3: What are the harms of screening for fracture risk or osteoporosis?

Three fair-quality, pragmatic RCTs (Table 1) were included: the Risk-stratified Osteoporosis Strategy Evaluation (ROSE) study (N = 34 229 randomized; N = 18 605 per protocol 1 analysis),^20–23 the Screening in the Community to Reduce Fractures in Older Women (SCOOP) study (N = 12 483),^24–27 and the Stichting Artsen Laboratorium en Trombosedienst Osteoporosis Study (SOS) (N = 11 032).^28,29 ROSE and SOS are new to this update. All 3 RCTs randomized European women (mean or median age 71 to 76 years) to screening vs usual care (ie, no systematic screening). Among those enrolled, the mean or median 10-year FRAX-estimated risk of hip fracture was 8.5% in SCOOP, 6.7% in ROSE, and 11.6% in SOS.^22,26,29 The proportion of participants with a prior fracture was 12.6% in ROSE, 22% in SCOOP, and 43% in SOS; however, there was variability in the definition and reporting of prior fractures between trials.^22,26,29 Three relevant SRs were also identified that analyzed these 3 trials and reported similar findings as this update.^30–33

Table 1.

Study Characteristics of Randomized Clinical Trials of Screening to Prevent Fracture (KQ1)

Study, Year	Recruitment Setting	Mean Age (SD)	% Female	N randomized	Intervention Groups (N Randomized)	Study Quality
ROSE 201720–23	Civic registries in southern Denmark	Median 71 [IQR 68, 76]	100	34 229a	Screening: FRAX without BMD assessment with invitation to DXA and VFA if 10-year FRAX MOF risk ≥ 15%; results sent to the participant and PCP with treatment recommendations based on national guidelines Routine care: no contact after completion of baseline data collection, usual care guided by PCP	Fair
SCOOP, 201824–27	General practice clinics in the UK	Screening: 75.5 (4.2) Routine Care: 75.5 (4.1)	100	12 483	Screening: FRAX without BMD assessment; if high risk based on 10-year FRAX hip risk ≥ age-specific threshold, then invitation to DXA; if below threshold, then letter sent to participants and PCPs confirming low-risk status; DXA results sent to participant and PCP with participant’s revised FRAX risk (including BMD information), age-specific treatment thresholds, and recommendation to discuss treatment if above threshold Routine care: letter informing PCP of patient’s participation in the study; usual care guided by PCP	Fair
SOS, 201928,29	General practice registries in the Netherlands; only women with 1 or more clinical risksb were recruited	75.0 (6.7)	100	11 032	Screening: FRAX without BMD assessment, DXA, VFA, fall risk assessment, and blood chemistries to exclude secondary osteoporosis; women with treatment indications based on results had referral to PCP for personalized treatment advice including medication, evaluation for secondary osteoporosis, fall prevention, and calcium/vitamin D supplementation; PCPs were provided group education on the study protocol and treatment options Routine care: wait list placement for screening intervention, usual care guided by PCP	Fair

^aThe intent-to-treat analysis was based on the number randomized, which occurred prior to any data collection about fracture risk. The authors also reported a per-protocol analysis for 18 605 participants. This per-protocol analysis is methodologically more similar to the designs of the SCOOP and SOS trial because only participants who completed baseline questionnaires about fracture risk were randomized to screening in those trials. Thus, the per-protocol analysis from the ROSE trial was used in the quantitative synthesis depicted in Figure 3.

^bClinical risk factors: previous fracture after aged 50, parental hip fracture, BMI < 19 kg/m², rheumatoid arthritis, menopause before 45 years, malabsorption syndrome, chronic liver disease, type I diabetes, immobility

BMD indicates bone mineral density; DXA, dual-energy X-ray absorptiometry; FRAX, Fracture Risk Assessment Tool; IQR, interquartile range; KQ, key question; N, number of participants; PCP, primary care provider; RCT, randomized clinical trial; ROSE, Risk-stratified Osteoporosis Strategy Evaluation; SCOOP, Screening in the Community to Reduce Fractures in Older Women; SD, standard deviation; SOS, Stichting Artsen Laboratorium enTrombosedienst Osteoporosis Study; UK, United Kingdom; VFA, vertebral fracture assessment.

Two RCTs (SCOOP²⁶ and ROSE²⁹) used 2-step screening with a FRAX risk assessment (without BMD) then invited those with a FRAX risk above a specified threshold for DXA. In contrast, SOS enrolled women known to have at least 1 clinical risk factor for osteoporosis and conducted DXA, vertebral fracture assessment, blood chemistries, falls risk assessment, and FRAX without BMD.²⁹ In all trials, DXA results and corresponding treatment recommendations were shared with the participant and primary care provider (PCP) who together made final decisions about starting treatment. The control intervention in all 3 trials was usual care as guided by the participant’s PCP. The study quality was fair in all trials because of contamination in control groups, poor to modest adherence in intervention groups, and lack of blinding, which was not feasible because of the pragmatic nature of the trials (detailed quality assessments are in Appendix D of the Full Report).

All 3 RCTs confirmed fracture outcomes through medical records or radiology reports and the primary outcomes were any fractures (SCOOP, SOS) and MOF (ROSE); hip fractures were secondary outcomes. The relative associations between screening and fracture or mortality outcomes are depicted in Figure 3. The pooled RR for hip fractures was 0.83 (95% CI, 0.73 to 0.93; 3 RCTs; 42 009 participants; I² = 0%); this corresponds to an absolute risk difference (ARD) of 5 fewer hip fractures per 1000 people screened (95% CI, from 7 fewer to 2 fewer). The pooled RR for MOF was 0.94 (95% CI, 0.88 to 0.99; 3 RCTs; 42 009 participants; I² = 0%), corresponding to an ARD of 6 fewer per 1000 screened (95% CI, from 12 fewer to 1 fewer). The pooled estimates for all fractures or osteoporotic fractures had similar ARDs but were not statistically significant. No significant association was observed for all-cause mortality; pooled RR, 0.99 (95% CI, 0.95 to 1.04; 3 RCTs; 57 633 participants; I² = 0%) corresponding to an ARD of 1 fewer death per 1000 screened (95% CI, from 5 fewer to 4 more).

Figure 3. Results of Randomized Clinical Trials of Screening vs Usual Care (KQ1).

Note: This analysis used the first per-protocol data from the ROSE trial for the fracture outcomes because these data reflect a similar study design as the intention-to-treat (ITT) data reported for SCOOP and SOS. See the full evidence report for a sensitivity analysis using the ITT data from the ROSE trial for the fracture outcomes. The data for mortality is the ITT population for ROSE because per-protocol data for ROSE was not reported.

^a SCOOP reported an outcome entitled “Osteoporotic Fractures,” which were defined as clinical fractures excluding hand, foot, skull, or cervical vertebrae. This definition differs from the definition of MOF used by the other 2 studies (hip, clinical vertebral, distal forearm, and humerus); as such, SCOOP data was included for both “Osteoporotic Fractures” and for “MOF” in this figure. The RR estimate for MOF without SCOOP included is 0.93 (95% CI, 0.86 to 1.00); Absolute Effect: 6 fewer (from 12 fewer to 0 fewer). It is also not clear that fractures associated with trauma were excluded from SCOOP.

CI indicates confidence interval; DL, DerSimonian & Laird estimator for pooling estimates; FRAX, Fracture Risk Assessment Tool; F/U, follow-up; ITT, inention to treat; KQ, key question; MOF, major osteoporotic fracture; N, number; PriorFx, prior fracture; ROSE, Risk-stratified Osteoporosis Strategy Evaluation; SCOOP, Screening in the Community to Reduce Fractures in Older Women; SOS, Stichting Artsen Laboratorium en Trombosedienst Osteoporosis Study; vs, versus; y, year.

Of the 3 RCTs included for KQ 1, only the SCOOP trial reported on harms of screening.^26,27 Anxiety was assessed using the Strait-Trait Anxiety Inventory-Short Form at repeated intervals over the 5-year study period.^26,27 Authors observed no difference in anxiety between screening participants (both those deemed low risk and those deemed high risk who were invited to DXA testing) and the control group participants (p=0.515).^26,27 One SR reported on overdiagnosis.^31,32 Based on the data reported in the SCOOP and SOS RCTs, the SR authors estimated the proportion of participants overdiagnosed ranged from 11.8% to 24.1%.^31,32

Accuracy of Screening Strategies

Key Question 2a: What is the predictive accuracy of risk assessment tools for identifying adults who are at increased risk for hip or major osteoporotic fractures?

Thirty cohort studies (published in 49 articles^34–82) and 6 SRs^31,83–87 reported on the accuracy (discrimination, calibration, or both) of 11 risk assessment instruments for predicting fracture (eTable 1 of the Supplement). All the SRs were good quality; however, authors of the SRs generally rated the included primary studies as poor quality. All of the primary studies included for KQ 2a in this update were rated as poor quality.

Two SRs^31,86 and 25 cohorts reported in 40 articles^{34–59,61–63,65,67,68,70,71,73,75,76,78–82} reported on calibration outcomes for 6 risk assessment models (FRAX, Fracture Risk Evaluation Model [FREM], FRC, Garvan, Osteoporosis Self-Assessment Tool[OST], QFracture) for the prediction of MOF, hip fracture, or both. Calibration results were heterogenous with no discernible patterns with respect to instrument, age, or sex.

Six SRs^{31,32,83–87} and 16 cohort studies published in 25 articles^{34,44,51,52,58–62,65,66,68–72,74,76–81} reported on the discriminative accuracy of 11 risk assessment models (EPIC, FRAX, FRC, FREM, Garvan, ORAI, OSIRIS, OST, QFracture, SCORE, Women’s Health Initiative Prediction Model) to predict MOF or hip fracture or both using primarily AUC. Findings were heterogenous (eFigures 1, 2, 3 in the Supplement), spanning a range considered poor accuracy (AUC 0.52) to very good accuracy (AUC 0.93); however, most AUCs were between 0.60 and 0.80. Sources of heterogeneity in AUC estimates include age and population evaluated along with variation in outcome definitions and analytic methods used by authors. Discrimination was largely similar in men and women. For risk assessment instruments with the option to include BMD (FRAX, FRC, Garvan), the predictive accuracy was improved when BMD was included compared to the instruments alone. Further, some instruments (FRAX, FRC, Garvan, QFracture) had higher accuracy for predicting hip fracture than for predicting MOF. Few studies reported sensitivity or specificity for specific risk thresholds. In 1 cohort of US women aged 50 to 64 years, a FRAX risk threshold of 9.3% had a 26% sensitivity and an 83% specificity to identify MOF.³⁴

Key Question 2b: What is the predictive accuracy of bone mineral density testing with dual X-ray absorptiometry at central skeletal sites for identifying adults who are at increased risk for hip or major osteoporotic fractures?

The accuracy of BMD measurement (typically at the femoral neck) for prediction of incident fractures was reported in 22 unique cohorts in 28 publications (eTable 2 and eFigure 4 in the Supplement).^{40,41,45–47,53,59,61–63,67,69,70,74,88–101} A third were poor-quality studies; most were in women, and the mean age of participants was 49 to 75 years. Four cohorts^40,62,90,99 reported at least 1 type of calibration outcome, but few authors reported detailed or consistent information. Substantial heterogeneity precluded quantitative synthesis of AUC, which ranged from 0.60 to 0.80 for BMD (treated as a continuous variable) for prediction of MOF (13 cohorts^{40,41,45–47,53,61,62,69,70,74,89,95,96,101}) and from 0.64 to 0.86 for hip fracture prediction (12 cohorts^{40,41,45,47,53,61,62,70,74,89,90,92,100,101}). Few studies reported sensitivity and specificity and thresholds varied.

Key Question 2c: What is the diagnostic accuracy of risk assessment tools for identifying adults with osteoporosis?

Forty-three unique cohorts (published in 54 articles.^{52,58,79,102–152}) reported on accuracy of risk assessment instruments for identifying osteoporosis (eTable 3 in the Supplement). More than half enrolled people with a mean age between 60 and 69 years and studies included women, men, or both. Differences in reference standards, risk assessment score thresholds, and study populations precluded quantitative synthesis. In women, AUCs ranged from 0.32 to 0.87 across 35 articles evaluating 11 instruments (eFigure 5 in the Supplement). Five articles reported results from 3 independent cohorts that retrospectively evaluated the accuracy of a FRAX MOF risk threshold of 8.4% or 9.3% in women aged 50 to 64 years with AUCs ranging from 0.55 to 0.62; sensitivity ranged from 5% to 49% and specificity ranged from 63% to 96%.^{58,102,127,128,137} In men, AUCs ranged from 0.62 to 0.94 across 18 articles evaluating 12 instruments (eFigure 6 in the Supplement). Three articles reported on accuracy among mixed populations of men and women for 3 instruments (eFigure 7 in the Supplement); findings were consistent with those reported for men and women separately.

Key Question 2d. What Is the Evidence to Determine Screening Intervals, and How Do These Intervals Vary by Baseline or Current Individual Fracture Risk?

Five cohort studies^153–157 including 3 new to this update^155–157 evaluated the accuracy of repeat BMD measurement to predict fracture risk at an interval of 4 to 8 years after initial BMD measurement. In 4 of the 5 studies, authors reported similar accuracy for models that used initial BMD, change in BMD, or both. In the fifth study, authors reported no association between change in spine, total hip, or femoral neck BMD on a repeat DXA at a mean of 4.1 years and MOF fracture (gradient of risk, HR 0.93, 95% CI 0.81 to 1.06 per SD increase in BMD at the femoral neck on repeat DXA).¹⁵⁵

Benefits of Pharmacotherapy

Key Question 4. What is the effectiveness of pharmacotherapy with selected FDA-approved medications on fracture incidence and fracture-related morbidity and mortality?

Twenty-one RCTs (reported in 27 articles^158–184) compared bisphosphonates (alendronate, ibandronate, risedronate, or zoledronic acid) with placebo and 6 RCTs (reported in 9 articles^185–193) compared denosumab with placebo. Two RCTs of alendronate,^182,183 2 RCTs of zoledronic acid,^175–1801 RCT of ibandronate,¹⁸¹ and 2 RCTs of denosumab^186,193 were new to this update. Three RCTs were good quality;^{158,161,173,175,176,178} the rest were fair quality. A summary of trial characteristics is in Table 2. One RCT of zoledronic acid¹⁵⁸ and 1 study of denosumab¹⁸⁶ were conducted exclusively in men; 3 studies (all evaluating bisphosphonates) included a small proportion of men. The remaining studies were conducted exclusively among postmenopausal women. T-score criteria for enrollment across studies varied, but only 6 required T-scores in the osteoporotic range. The rest enrolled participants with T-scores spanning the range considered low bone mass and osteoporosis or low bone mass only. Detailed trial characteristics, study quality assessments, and results are in Appendix D of the Full Report.

Table 2.

Randomized, Placebo-Controlled Trials of Treatment for Osteoporosis (KQs 4 and 5)

Author, Year	Study Quality	Total N	% Female	Mean Age (SD)	Race/Ethnicity	% with Prior Fracturea	T-Score Inclusion Criteria	Dose and Duration	Key Question
Alendronate
Adachi et al, 2009 205	Fair	438	100%	65.5 (NR)	89% White, 8% Hispanic, 3% Asian, 1% Black	6.8%	< −2.0	10 mg per day; 3 months	KQ 5
Ascott-Evans et al, 2003159	Fair	144	100%	57.3 (6.6)	91.7% White, 8.3% other	0%	LS < −1.5 and > −3.5	10 mg per day;1 year	KQ4, KQ5
Bell et al, 2002182	Fair	65	100%	66 (NR)	100% African-American	NR	LS < −1.75	10 mg per day; 2 years	KQ 4, KQ 5
Bone et al, 1997183	Fair	359	100%	71 (NR)	97% White	34% to 42%	≤ −2.0	1, 2.5, or 5 mg per day; 2 years	KQ 4, KQ 5
Chesnut et al, 1995160	Fair	188	100%	62.9 (6.1)	97.9% White, 2.1% Asian	0%	NR; mean T-score −1.1	Variousb; 2 years	KQ 4, KQ 5
Cryer et al, 2005206	Fair	454	100%	65 (10)	91% White, 2% Black, 1% Asian, 5% Hispanic. 1% Native American, 1% other	NR	Any site < −2.0 and > −3.5	70 mg weekly; 6 months	KQ 5
Cummings et al, 1998161 Bauer et al, 2000197 Cummings et al, 2007198 Quandt et al, 2005173 FIT	Good	4432c	100%	67.6 (6.2)	97% White	0%c	FN < −1.6	5 mg per day for 2 years then 10 mg per day for 1 year; 3 years	KQ 4, KQ 5
Devogelaer et al, 1996211	Fair	516	100%	62 (NR)	NR	NR	LS ≤ −2.5	5, 10, 20d mg per day; 3 years	KQ5
Eisman et al, 2004207	Good	449	93–96%	63.6 (NR)	65.7% White, 18% Asian, 12% Hispanic, 5% other	NR	NR; mean T-score NR	70 mg weekly; 3 months	KQ 5
Greenspan et al, 2002208	Fair	450	92%	67 (NR)	96% White	NR	NR; mean T-score NR	70 mg weekly; 3 months	KQ 5
Greenspan et al, 2003209	Good	186	100%	71.5 (NR)	NR	0%	NR; mean T-score −1.7	10 mg per day; 3 years	KQ 5
Hosking et al, 2003172	Fair	549e	100%	69 (NR)	99.5% Caucasian	48.5%	LS or TH < −2.5 or both < −2.0	70 mg weekly; 1 year	KQ 4, KQ 5
Johnell et al, 2002202	Fair	331	100%	63.6 (NR)	95% White	NR	FN < −2.0	10 mg per day; 1 year	KQ 5
Liberman et al, 1995162	Fair	994	100%	64 (NR)	87.4% White, 0.4% Black, 12.2% other	21%	LS < −2.5	5 or 10 mg per day; 3 years 20 mg per day for 2 years followed by 5 mg/day for 1 year	KQ 4, KQ 5
Pols et al, 1999165	Fair	1908	100%	62.8 (7.5)	94% White	NR	NR; mean T-score	10 mg per day; 1 year	KQ 4, KQ 5
Tucci et al, 1996174	Fair	478	100%	64 (NR)	91% White, 8% Asian	NR	LS < −2.5	5 mg, 10 mg, or 20 mg per day for 2 years followed by 5 mg per day; 3 years	KQ 4, KQ 5
Ibandronate
Chapurlat et al, 2013195	Fair	148	100%	62.7 (5.0)	NR	NR	LS or TH < −1.0 and > −2.5	150 mg per month; 2 years	KQ 5
McClung et al, 2009181	Fair	160	100%	53 (NR)	NR	0%	LS < −1.0 and > −2.5 with TH or FN > −2.5	150 mg per month; 1 year	KQ 4, KQ 5
McClung et al, 2004199	Fair	653	100%	58.2 (8.6)	NR	0%	LS < −1.0 and > −2.5	0.5 mg, 1.0 mg, or 2.5 mg per day; 2 years	KQ 5
Ravn et al, 1996169	Fair	180	100%	65 (NR)	100% White	0%	NR; mean T-score −1.7	0.25 mg, 0.50 mg, 1.0 mg, 2.5 mg, or 5.0 mg per day; 1 year	KQ 4, KQ 5
Reginster et al, 2005171	Fair	144	100%	65.7 (NR)	NR	NR	NR; mean T-score −0.3 to −1.9	Variousf;3 months	KQ 4, KQ 5
Riis et al, 2001170	Fair	240	100%	66.8 (4.9)	NR	NR	LS or FN < −2.5	2.5 mg per day or intermittent cyclic dose; 2 years	KQ 4, KQ 5
Tanko et al, 2003200	Fair	630	100%	55 (NR)	NR	0%	LS ≥ −2.5	5 mg, 10 mg, or 20 mg weekly; 2 years	KQ 5
Thiebaud et al, 1997201	Fair	126	100%	64 (NR)	NR	0%	LS < −2.5	0.25, 0.5 mg, 1.0, or 2.0 mg every 3 months; 1 year	KQ 5
Risedronate
Bala et al, 2014216	Fair	324	100%	53–61	NR	NR	LS or TH < −1.0 and > −2.5	35 mg weekly; 1 year	KQ 5
Hosking et al, 2003172	Fair	549e	100%	69 (NR)	99.5% Caucasian	48.5%	LS or TH < −2.5 or both < −2.0	5 mg daily; 3 months	KQ 4, KQ 5
McClung et al, 2001163	Fair	9331	100%	NR, all 70 years or older	98% White	39% to 44%	FN < −4 or < −3 with risk factor for hip fracture	2.5 or 5 mg per day; years	KQ 4, KQ 5
Mortensen et al, 1998164	Fair	111	100%	52.1 (3.9)	100% White	0%	Z-score > −2.0; mean T-score −1.1	5 mg cyclic or 5 mg per day; 2 years	KQ 4, KQ 5
Shiraki et al, 2003203	Fair	211	99%	60.3 (NR)	100% Japanese	Mean prevalent vertebral fractures 0.3 (SD 0.8)	LS < −2.5 without vertebral fracture; < −1.5 with vertebral fracture	1 mg, 2.5 mg, or 5 mg per day; 8 months	KQ 5
Valimaki et al, 2007167	Fair	170	100%	65.9 (6.8)	100% White	NR	LS > −2.5 and < −1 and proximal femur ≤ −1	5 mg per day; 2 years	KQ 4, KQ 5
Zoledronic Acid
Boonen et al, 2012158	Good	1199	0%	Median 66	94% White, 1% Black, 1% Asian, 0.5% other	32%	TH or FN ≤ −1.5	5 mg every year; 2 years	KQ 4, KQ 5
Grey et al, 2010168 Grey et al, 2009184	Fair	50	100%	62 (8)	NR	42%	LS or TH < −1 and > −2	5 mg; single dose with 3 year follow-up	KQ 5
Grey et al, 2012179 Grey et al, 2014180 Grey et al, 2017177	Fair	180	100%	66 (9)	NR	14% to 21%	LS or TH < −1 and > −2.5	1 mg, 2.5 mg, 5 mg; single dose	KQ 4, KQ 5
McClung et al, 2009194	Fair	581	100%	59.6 to 60.5	NR	0%	LS −1.0 and −2.5 and FN > −2.5	5 mg single dose or 5 mg yearly for 2 years; 2 years	KQ 5
Reid et al, 2002166	Fair	351	100%	65 (7)	95% White	0%	LS < −2.0	Variousg;1 year	KQ 4, KQ 5
Reid et al, 2018176 Reid et al, 2019175 Reid et al, 2020210 Reid et al, 2021178	Good	2000	100%	71 (5.1)	95% European, 0.02% Maori, 0.01% Pacific Islander, 0.02% East Asian, 0.005% Indian, 0.002% other	23.7%	TH or FN −1.0 to −2.5	5 mg every 18 months; 6 years	KQ 4, KQ 5
Denosumab
Bone et al, 2008191	Fair	332	100%	59.4 (7.5)	NR	0%	LS or TH between −1 and −2.5	60 mg every 6 months; 3 years	KQ 4, KQ 5
Cummings et al, 2009189 Watts et al, 2012196 Simon et al, 2013185 McCloskey et al, 2012187 Palacios et al, 2015192 FREEDOM	Fair	7,808	100%	72.3 (5.2)	NR	50%	LS or TH < −2.5 but > −4.0	60 mg every 6 months; 3 years	KQ 4, KQ 5
Koh et al, 2016193	Fair	135	100%	67.0 (4.9)	NR	23% to 30%	TH or LS < −2.5 and ≥ −4.0	60 mg; single dose with 6 month follow-up	KQ 4, KQ 5
Lewiecki et al, 2007190 McClung et al, 2006204	Fair	365	100%	62.5 (8.1)	86.2% White, 9.5% Hispanic, 2.9% Black, 1.5% other	0%	LS −1.8 to −4.0 or FN −1.8 to −3.5	Varioush;2 years	KQ 4, KQ 5
Nakamura et al, 2012188	Fair	226	100%	65.1 (6.8)	100% Japanese	34%	LS −2.5 to −4.0 or FN or TH −2.5 to −3.5	Variousi;1 year	KQ4, KQ5
Orwoll et al, 2012186 ADAMO	Fair	242	0%	65.0 (9.8)	94.2% White	39.3%	LS or FN −2.0 to −3.5j;or LS or FN −1.0 to −3.5j with prior MOF	60 mg every 6 months; 2 years	KQ4, KQ5

^aStudies define this in varying ways: any fracture, fracture after age 50, fragility fracture, vertebral fracture only.

^b5 mg/day or 10 mg/day or 40 mg/day for 3 months then 2.5 mg/day for 21 months; 20 mg/day for 1 year then placebo for 1 year; 40 mg/day for 1 year then placebo for 1 year.

^cOnly the portion of the enrolled population without prior vertebral fracture was used in this review.

^dDosage was 20 mg for first 2 years and lowered to 5 mg in the final year.

^eIncludes the alendronate, risedronate, and placebo arms.

^f50 mg per month; 50 mg for the first month then 100 mg for months 2–3; 100 mg per month; 150 mg per month.

^g0.25 mg every 3 months, 0.5 mg every 3 months, 1 mg every 3 months, 4 mg every 1 year, 2 mg every 6 months.

^h6 mg, 14 mg, or 30 mg every 3 months; 14 mg, 60 mg, 100 mg, or 210 mg every 6 months.

ⁱ14 mg, 60 mg, or 100 mg every 6 months.

^jT-scores based on male reference range.

FIT indicates Fracture Intervention Trial; FN, femoral neck; FREEDOM, Fracture Reduction Evaluation of Denosumab in Osteoporosis Every 6 Months; KQ, key question; LS, lumbar spine; MOF, major osteoporotic fracture; N, number; NR, not reported; SD, standard deviation; TH, total hip.

Pooled results from included trials reporting vertebral fractures (clinical, radiographic, or both), nonvertebral fractures, hip fractures, and mortality are in Figure 4. Pooled RRs ranged from 0.33 to 0.81 across drugs and outcomes with corresponding ARDs from 3 to 44 fewer events (fractures or deaths) per 1000 people treated. Findings from sensitivity analyses were consistent for each outcome when alternative pooling methods or dosages other than FDA-approved dosages were used (Appendix E.4 in the Full Report). In the single trial conducted among men (N = 1199 with T-score less than −1.5 based on device specific reference values), authors reported a reduced risk of radiographic vertebral fractures (1.5% vs 4.6%; RR, 0.33 [95% CI, 0.16 to 0.70]) but no significant difference in nonvertebral fractures (0.9% vs 1.3%; RR 0.65 [95% CI, 0.21 to 1.97]) compared with placebo.¹⁵⁸

Figure 4. Results of Randomized, Placebo-Controlled Trials of Treatment for Osteoporosis (KQs 4 and 5).

^a Although multiple studies reported, evidence base is dominated by 1 large (N = 7808) study.

^b Sensitivity analysis was conducted limiting to studies reporting clinical vertebral fractures ( 4 studies) and the pooled RR was 0.44 (95% CI, 0.24 to 0.79; 2373 participants, I² = 0%).

AE indicates adverse event; ARD, absolute risk difference; CI, confidence interval; GI, gastrointestinal; KQ, key question; N, number; RR, relative risk.

Harms of Pharmacotherapy

Key Question 5. What are the harms associated with selected FDA-approved medications?

Forty RCTs (reported in 48 articles^{158–172,174,176,177,179–184,186,188–191,193–211}) compared bisphosphonates or denosumab with placebo and reported harm outcomes. In addition, 3 controlled cohort studies evaluated bisphosphonates compared with placebo.^212–214 Five RCTs were good quality^{158,161,176,197,198,207,209,210}; the rest of the RCTs and the controlled cohort studies were fair quality. A summary of RCT characteristics is in Table 2 and pooled findings from included trials reporting discontinuations due to adverse events (AE), serious AE, or gastrointestinal AE are in Figure 4. Across these outcomes, pooled RRs ranged from 0.97 to 2.18 with corresponding ARDs from 6 fewer to 14 more per 1000 people treated with no statistically significant associations observed. For bisphosphonates, 8 RCTs^{158,161,168,176,179,194,203,209} reported 1 or more cardiovascular outcomes (eg, incidence of myocardial infarction, atrial fibrillation); generally these events were rare and estimates were imprecise (details in the Full Report). For denosumab, 3 RCTs reported additional harm outcomes related to skin disease and infection and with 1 exception (incidence of eczema in 1 RCT¹⁸⁹), no associations were observed.^189–191

For bisphosphonates, 5 RCTs reported 0 cases of osteonecrosis of the jaw^{158,168,176,179,194} and no RCTs reported on the rare outcome of atypical femur fracture. For denosumab, 3 RCTs^186,189,193 reported 0 cases of osteonecrosis of the jaw. Two RCTs^186,193 reported on the rare outcome of atypical femur fracture and both reported 0 events. Additional information about these rare outcomes from study designs not eligible for inclusion were addressed as a Contextual Question (Appendix F.3 in the Full Report). No studies that were included for KQ 5 had study designs sufficient to evaluate rebound vertebral fractures after denosumab discontinuation. Findings related to rebound vertebral fractures from studies not eligible for inclusion in this update were addressed as a Contextual Question (Appendix F.4 in the Full Report).

Three fair-quality cohort studies set in Denmark,²¹⁴ Sweden and Denmark,²¹³ and South Korea²¹² addressed potential harms of bisphosphonate use. Two studies were limited to new users;^212,214 the third study provided sensitivity analyses for a treatment-naïve cohort.²¹³ The studies predominantly (86%²¹³ and 91%²¹²) or solely comprised women.²¹⁴ One study was limited to zoledronic acid,²¹³ a second to alendronate,²¹⁴ and the third included all bisphosphonates (which may have included non-FDA-approved bisphosphonates).²¹² Detailed study characteristics, quality assessments, and results are in Appendix D Tables 7 and 17 of the Full Report. In brief, 2 of the studies reported an increased risk for atypical femur fractures with bisphosphonate use compared with nonusers (aHR, 2.46 [95% CI, 1.17 to 5.15], N = NR²¹³ and aHR, 1.53 [95% CI, 1.36 to 1.73], N = 696 859²¹²). However, neither study controlled for all known confounders such as smoking, BMI, or alcohol use.

DISCUSSION

The SOE by KQ is presented in Table 3. Compared with the prior review,² our certainty related to the direct benefits of screening has increased because of new evidence for KQ1. In contrast, the evidence remains insufficient for harms of screening (KQ3). Based on the screening strategies evaluated, the SOE was rated as moderate for a small absolute benefit on MOF and hip fractures, low SOE for a small absolute benefit on osteoporotic fractures, and low SOE for no effect on mortality; however, no direct evidence for BMD screening with DXA alone is available. The 3 studies included for KQ1 were pragmatic trials conducted among older European women (median aged 71 to 76 years) at relatively high risk for fracture (10 year estimated FRAX risks at baseline ranged from 6.7% to 11.6%). The proportion of eligible persons who participated was low (about one-third) in 1 trial²⁶ with evidence of selection bias toward healthy individuals, and the receipt of the screening intervention was suboptimal in the other 2 trials (55%²² and 76%²⁹). These trials were underpowered because the observed proportion of women with treatment indications and who adhered to treatment were lower than expected and because of contamination in control arms from secular trends in screening and treatment. For these reasons, the estimates of benefits probably represent the lower bounds of screening efficacy. Yet, these findings may reflect the real-world effectiveness of a systematic screening program. Although these estimates represent the lower bounds of efficacy, it is not entirely clear that the findings are applicable to populations with lower fracture risk or US settings given the use of country-specific FRAX prediction models and the thresholds for action (further DXA testing or treatment) used in these trials.

Table 3.

Summary of Evidence on Screening for Osteoporosis to Prevent Fracture

Key Question	Intervention or Test/ Outcome	No. of Studies (No. of Participants)	Summary of Findings	Consistency and Precision	Limitations	Strength of Evidence	Applicability
1 Benefits of Screening	Fractures	3 RCTs),20–29 (42 009 using ROSE per protocol 1 population) 3 SRs30–33	Hip fractures pooled RR, 0.83 (95% CI, 0.73 to 0.93); ARD, 5 fewer per 1000 (95% CI, from 7 fewer to 2 fewer) MOF pooled RR, 0.94 (95% CI, 0.88 to 0.99); ARD, 6 fewer per 1000 (95% CI, from 12 fewer to 1 fewer) Osteoporotic fractures pooled RR, 0.95 (95% CI, 0.91 to 1.01); ARD, 6 fewer per 1000 (95% CI, from 11 fewer to 1 more) Estimates from SRs consistent.	Consistent, Precise for hip and MOF, Imprecise for osteoporotic	Modest screening uptake and adherence to treatment; contamination in control groups; follow-up for only 3.7 to 5 years	Moderatea for benefit on MOF and hip fracture; Lowb for benefit on osteoporotic fractures	Two-stage screening used by 2 studies; European women 60 years or older at high baseline fracture risk; extensive screening battery (imaging, labs, falls assessment) used in 1 study
	Mortality	3 RCTs20–29 (57 633) 1 SR31,32	Pooled RR, 0.99 (95% CI, 0.95 to 1.04) ARD, 1 fewer per 1,000 (95% CI, from 5 fewer to 4 more) Estimates from SR consistent.	Consistent, imprecise	Same as above	Lowb for no effect	Same as above
2a Predictive Accuracy of Risk Assessment Instruments	Calibration (MOF and Hip Fracture)	Two SRs31,32,86 and 25 cohorts reported in 40 articles34–51,53–59,61–63,65,68,70,71,73,75,76,78–82 (Unable to estimate precisely due to overlap in reporting for some cohorts)	Reported for 6 instruments: FRAX, FREM, FRC, Garvan, OST, QFracture FRAX (28 articles from 20 unique cohorts): reasonably calibrated in some cohorts and poorly calibrated in others Too few studies reported calibration for instruments other than FRAX.	Varied by instrument	All studies high risk of bias	Lowc for FRAX for poor to modest calibration Insufficientd for FRC, FREM, Garvan, OST, QFracture	Studies included postmenopausal women and men
2a Predictive Accuracy of Risk Assessment Instruments	Discrimination (MOF and Hip Fracture)	Four SRs83–86 16 cohorts published in 25 articles34,44,51,52,58–62,65,66,68–72,74,76–81 (Unable to estimate precisely due to overlap in reporting for some cohorts)	Reported for 11 instruments: EPIC, FRAX, FRC, FREM, Garvan Fracture Risk Calculator, ORAI, OSIRIS, OST, QFracture, SCORE, WHI Prediction Model AUC range Younger women (<65 years): 0.52 to 0.71 Women: 0.63 to 0.89 Men: 0.63 to 0.93 Mixed sex: 0.61 to 0.88 FRAX, FRC, and Garvan instruments with BMD had higher AUCs compared to same instrument without BMD. AUCs higher for prediction of hip fracture compared to MOF for FRAX, FRC, QFracture, and Garvan.	Varied by instrument	All studies high risk of bias for development cohorts and for external validation cohorts	Lowc for FRAX, FRC, Garvan, QFracture for poor to modest discrimination Insufficientd for EPIC, FREM, OST, SCORE, WHI	Studies included postmenopausal women and men, but not for all instruments
2b Predictive Accuracy of BMD	Calibration (MOF and Hip Fracture)	4 articles from 4 unique cohorts40,62,90,99 (18,145)	Inconsistent calibration measures reported across studies; calibration poor in some studies and good in others for prediction of MOF or hip fracture.	Inconsistent; unable to judge precision	Not the primary aim of any study; not enough fracture events in some studies, particularly for hip fractures	Insufficiente	Cohorts include both men and women; persons with known osteoporosis or on treatment excluded from some cohorts; BMD typically measured at FN
2b Predictive Accuracy of BMD	Discrimination (MOF and Hip Fracture)	18 articles from 16 unique cohorts40,41,45–47,53,61,62,69,70,74,89,90,92,95,96,100,101(101 446)	AUC range MOF:0.60 to 0.80 (13 cohorts; 15 estimates) Hip: 0.64 to 0.86 (12 cohorts; 14 estimates) Threshold T-score < −2.5 Sn MOF: 17.5% to 51.3% (5 studies) Sn Hip: 25.0% to 66.7% (5 studies) Sp MOF: 70.9% to 95.4% (3 studies) Sp Hip: 88.6% to 94.0% (4 studies)	Inconsistent, precise	10 analyses were high ROB; predictive accuracy of BMD not the primary aim of any study	Lowc for poor to modest dis-crimination	Same as above
2c Diagnostic Accuracy (Women and Men)	FRAX/ Discrimination	MOF risk 15 studies from 12 unique cohorts58,79,102,103,107,127,128,137,140,142,143,145–147,149 (37 756/85% women) Hip Fx risk 3 studies from 3 unique cohorts103,142,147 (1710/52% women)	MOF (9.3% or 8.4% risk threshold) Women Aged 50 to 64 years (3 estimates) AUC: 0.55 to 0.62 Sn: 5% to 49%; Sp: 63% to 96% Men (2 estimates) AUC: 0.62 to 0.79 Sn: 39% to 59%; Sp: 59% to 89% MOF (> 20% risk threshold) Women 60 years or older (1 estimate) AUC: 0.71 (95% CI, 0.60 to 0.82) Sn: 17%; Sp: 96% Men (1 estimate) Sn: 0%; Sp: 99% Mixed sex (1 estimate) AUC: 0.76 (95% CI, 0.71 to 0.81) MOF (various thresholds or no threshold) Women Aged 50 to 64 years (2 estimates) AUC: 0.64 to 0.72 Men (1 estimate) AUC: 0.62 Mixed sex (1 estimate) AUC:0.68 (95% CI, 0.63 to 0.72) Hip (>3% risk threshold) Women 60 years or older (1 estimate) AUC: 0.75 (95% CI, 0.65 to 0.86) Sn: 83%; Sp: 54% Men (1 estimate) AUC 0.86 (95% CI, 0.73 to 0.98) Sn: 80%; Sp: 71% Mixed sex (1 estimate) AUC: 0.70 (95% CI, 0.64 to 0.75)	Inconsistent, precise	Heterogeneity in BMD sites measured; all but 1 fair quality because of unclear methods for patient selection and risk for selection bias, lack of blinding of index or reference test results, unclear BMD reference range used for T-score, unclear interval between risk assessment and BMD measurement	Lowc for poor to modest dis-crimination	Men and postmenopausal women from community or clinic-based populations; FRAX risk assessment without BMD input into calculation; some studies used EHR data to determine FRAX risks
2c Diagnostic Accuracy (Women and Men)	OST/ Discrimination	31 studies from 29 cohorts 52,79,102,103,105–109,113,116–118,122,124–126,128,129,131,133,136–139,143–145,148,150,151 (80 592/82% women)	AUC (95% CI) or range: Women (20 estimates): 0.32 to 0.89 Women aged 50 to 64 years (3 estimates): 0.63 to 0.75 Men (10 estimates): 0.63 to 0.89 Mixed sex (1 estimate): 0.76 (0.71 to 0.82) At a score threshold of < 2: Women (11 estimates) Sn: 53% to 95%; Sp: 37% to 72% Women ages 50 to 64 (3 estimates) Sn: 56% to 79%; Sp: 56% to 70% Men (7 estimates) Sn: 62% to 89%; Sp: 36% to 74%	Inconsistent, precise	All but 1 fair quality; similar limitations as for FRAX above	Lowc for poor to modest dis-crimination	Men and postmenopausal women from community or clinic-based populations
2c Diagnostic Accuracy (Women)	Other risk assessments/ Discrimination	29 studies from 26 cohorts102,106,109,110,114–119,121–123,125,128,130–139,142,147,150,152 (30 621)	AUC range 0.32 to 0.87 (25 estimates) Across various thresholds: Sn range: 28% to 100% (24 estimates) Sp range: 5% to 100% (24 estimates)	Inconsistent; precision varies by instrument	All fair quality; similar limitations as for FRAX above	Lowc for poor to modest dis-crimination (ABONE, NOF, ORAI, OSIRIS, OSTA, SCORE) Insufficiente (AMMEB, Garvan FRC, SOFSURF)	Postmenopausal women from community and clinic-based populations
2c Diagnostic Accuracy (Men)	Other risk assessments/ Discrimination	21 studies 103–105,108,111,112,120,124,126,129,140–146,148,149,151 (24 258)	AUC range 0.64 to 0.88 in the studies exclusively enrolling men and evaluating instruments developed specifically for men; AUC range 0.62 to 0.94 from the male population component of the studies with mixed populations	Inconsistent; precision varies by instrument	All but 1 study fair quality; similar limitations as for FRAX above	Lowc for poor to modest dis-crimination (FRAX, MORES, MOST, OST, OSTA) Insufficiente (ABONE, Garvan FRC, MSCORE, ORAI, OSIRIS, SCORE, VA-FARA)	Men mostly from clinic-based populations
2d Repeat Screening	BMD at baseline and repeat BMD	5 studies153–157 (19 957)	Predictive accuracy of repeat BMD at 4 to 8 years after initial BMD was similar to predictive accuracy of initial BMD for predicting MOF and hip fractures over follow-up of 8 to 11 years after repeat BMD	Consistent, precise	2 studies were poor quality; 3 were fair quality; indirect evidence	Moderatef for no added value of repeat DXA	1 study exclusively in men;1 study with 40% men; mean aged 60 to 75 years across studies
3 Harms of screening	Anxiety	1 RCT26 (12 483)	No difference in anxiety between screening and control participants over 5 years (P = .515)	Single study, consistency unknown; precision unknown	Fair-quality pragmatic trial; modest uptake and adherence of intervention	Insufficientd	Two-stage screening approach in UK women aged 70 to 85 years
	Overdiagnosis	1 SR31 (NA)	Based on data from 2 included RCTs, overdiagnosis estimated to range from 11.8% to 24.1%	Single review, consistency unknown; precision unknown	Good quality SR; however, included RCTs are fair quality; method for estimating overdiagnosis for being labeled as “high risk” is evolving	Insufficientg	Two-stage screening in UK women aged 70 to 85 years in 1 study; Dutch women 60 years or older at high baseline fracture risk and extensive screening (imaging, labs, falls assessment) in other study
4 Benefits of treatment	Bis-phosphonates Vertebral Fx (clinical and radiographic)	10 RCTs158–162,164,166,167,176,183 (9015)	Pooled RR, 0.51 (95% CI, 0.39 to 0.66); ARD, 18 fewer per 1000 (95% CI, from 23 fewer to 13 fewer)	Consistent, precise	Most studies fair quality; evidence dominated by 3 larger studies; 5 studies had 0 events in at least 1 study arm	Moderatea for benefit	Only 1 study in men; the rest were in mostly White postmenopausal women with low bone mass or osteoporosis
	Bis-phosphonates Nonvertebral Fx	13 RCTs158,159,161–167,174,176,179,183 (20 929)	Pooled RR, 0.81 (95% CI, 0.74 to 0.88); ARD, 28 fewer per 1,000 (95% CI, from 38 fewer to 18 fewer)	Consistent, precise	Most studies fair quality, evidence dominated by 6 larger studies; 2 studies had 0 events in at least 1 arm	Moderatea for benefit	Only 1 study in men; the rest were in mostly White postmenopausal women with low bone mass or osteoporosis
	Bis-phosphonates Hip Fx	6 RCTs 161–165,176 (12 055)	Pooled RR, 0.67 (95% CI, 0.45 to 1.0); ARD, 3 fewer per 1000 (95% CI, from 5 fewer to 0 fewer)	Consistent, imprecise	Most studies fair quality; none were powered to evaluate hip fractures; 1 study had 0 events in at least 1 arm	Lowb for benefit	All studies in mostly White postmenopausal women with low bone mass or osteoporosis
	Bis-phosphonates Mortality	6 RCTs158,169–171,176,181 (3714)	Pooled RR, 0.71 (95% CI, 0.49 to 1.05); ARD, 10 fewer per 1000 (95% CI, from 17 fewer to 2 more)	Consistent, imprecise	Most studies fair quality; none were powered to evaluate mortality; 3 studies with 0 events in at least 1 arm	Lowb for benefit	Only 1 study in men, the rest were in mostly White postmenopausal women with low bone mass or osteoporosis
	Denosumab Vertebral Fx	4 RCTs186,188,189,191 (8179)	Evidence base dominated by FREEDOM study (n = 7808 women), RR, 0.32 (95% CI, 0.26 to 0.41), ARD, 48 fewer per 1000 participants (95% CI, from 52 fewer to 42 fewer) All other studies with 0 to 1 events per arm; pooled RR across all 4 RCTs, 0.33 (95% CI, 0.26 to 0.41); ARD, 44 fewer per 1000 persons (95% CI, from 49 fewer to 39 fewer)	Consistent, precise	All studies fair quality; evidence dominated by one study; outcome included both clinical and asymptomatic morphometric fractures	Moderatea for benefit	Postmenopausal women with osteoporosis or low bone mass; 1 study was only in men but had only 1 fracture event
	Denosumab Nonvertebral Fx	3 RCTs 186,189,191 (8382)	Evidence base dominated by FREEDOM study (n = 7808 women), 6.1% vs 7.5%; RR, 0.80 (95% CI, 0.67 to 0.95); ARD, 15 fewer per 1000 participants (95% CI, from 24 fewer to 4 fewer) Across all 3 RCTs, pooled RR, 0.80 (95% CI, 0.68 to 0.94); ARD, 14 fewer per 1000 (95% CI, from 23 fewer to 4 fewer)	Consistent, imprecise	Fair quality studies; evidence dominated by 1 large study	Lowb for benefit	Postmenopausal women with osteoporosis or low bone mass; 1 trial was only in men but had only 3 events
	Densoumab Hip Fx	2 RCTs186,189 (8050)	Evidence base dominated by FREEDOM study (n = 7808 women), 0.7% vs 1.1%; RR, 0.60 (95% CI, 0.37 to 0.97); ARD, 4 fewer per 1000 (95% CI, from 7 fewer to 0 fewer) 0 events in the other trial involving 242 men Across both studies, pooled RR, 0.61 (95% CI, 0.38 to 0.99), ARD 4 fewer per 1000 from 7 fewer to 0 fewer)	Consistent, imprecise	Fair quality; large trial with uncertainties in randomization/ allocation concealment, blinding, and attrition; no events in the other trial	Lowb for benefit	Postmenopausal women with osteoporosis or low bone mass; smaller trial was only in men but had no fracture events
	Denosumab Mortality	5 RCTs186,189–191,193(8828)	Pooled RR, 0.79 (95% CI, 0.58 to 1.07); ARD, 4 fewer per 1000 (95% CI, from 9 fewer to 1 more)	Consistent, imprecise	Fair quality; some un-certainties in randomization for 3 studies, allocation concealment in 4 studies, and attrition and blinding in 2 studies	Lowb for benefit	Postmenopausal women with osteoporosis or low bone mass; 1 trial only in men but had only 2 events
5 Harms of treatment	Bis-phosphonates Dis-continuations due to AEs	27 RCTs159–167,169,171,172,174,179,181–183,195,199–202,205–208,211 (18 617)	Based on 24 RCTs: Pooled RR, 1.00 (95% CI, 0.92 to 1.08); ARD, 0 fewer per 1000 (95% CI, from 9 fewer to 9 more)	Consistent, precise	Most studies fair quality, none powered for this outcome	Moderatea for no effect	Mostly White postmenopausal women with low bone mass or osteoporosis
	Bis-phosphonates SAEs	22 RCTs158,163,165–167,169,171,172,174,179,181,182,194,195,199–201,203,205,206,208,211 (13 878)	Based on 21 RCT comparisons: Pooled RR, 0.97 (95% CI, 0.91 to 1.04); ARD, 6 fewer per 1000 (95% CI, from 18 fewer to 8 more)	Consistent, precise	Most studies fair quality, none powered for this outcome, not long enough to detect rare harms	Moderateb for no effect	Only 1 study exclusively in men, the rest were in mostly White postmenopausal women with low bone mass or osteoporosis
	Bis-phosphonates Upper gastro-intestinal AE	26 RCTs159,162–165,167,169,171,172,174,176,179,181,182,197,199–203,205–209,211 (22 280)	Pooled RR, 1.02 (95% CI, 0.98 to 1.06); ARD, 5 more per 1000 (95% CI, from 5 fewer to 16 more)	Consistent, precise	Most studies fair quality, none powered for this outcome	Moderatea for no effect	Mostly White postmenopausal women with low bone mass or osteoporosis
	Denosumab Discontinuations due to AEs	5 RCTs186,189–191,193 (8826)	Pooled RR, 1.16 (95% CI, 0.87 to 1.54); ARD, 3 more per 1000 (95% CI, from 3 fewer to 11 more)	Consistent, imprecise	Fair quality; some un-certainties in randomization for 3 studies, allocation concealment in 4 studies, and attrition and blinding in 2 studies.	Lowb for no effect	Postmenopausal women with osteoporosis or low bone mass
	Denosumab Serious AEs	6 RCTs186,188–191,193 (8934)	Pooled RR, 1.04 (95% CI, 0.97 to 1.12); ARD, 9 more per 1000 (95% CI, from 7 fewer to 28 more)	Consistent, imprecise	Fair quality; some uncertainty for allocation concealment in all studies, randomization in 4 studies, and attrition and masking in 2 studies; not large enough or long enough to detect rare harms	Lowb for no effect	Postmenopausal women with osteoporosis or low bone mass
	Denosumab Upper GI AEs	4 RCTs 188–191,193 (932)	Pooled RR, 2.18 (95% CI, 0.74 to 6.46); ARD, 14 more per 1000 (95% CI, from 3 fewer to 66 more)	Consistent, imprecise	Fair quality; some uncertainty for allocation concealment in all studies, randomization in 3 studies, and attrition and masking in 1 study	Lowb for harm	Postmenopausal women with osteoporosis or low bone mass

^aRated down 1 level for study limitations.

^bRated down 1 level for imprecision and 1 level for study limitations.

^cRated down 1 level for inconsistency and 1 level for study limitations.

^dNot enough data to evaluate SOE

^eDowngraded 1 level for study limitations, 1 level for inconsistency, and 1 level for imprecision.

^fDowngraded one level for study limitations, including indirectness as these study designs did not directly compare a strategy of repeat screening with single screening.

^gNot enough data to evaluate SOE and indirect evidence based on extrapolations.

ABONE indicates Age, Body Size, No Estrogen instrument; AE, adverse event; AMMEB, Age, Menopause, Menarche, Body Mass Index; ARD, absolute risk difference; AUC, area under the curve; BMD, bone mineral density; CI, confidence interval; DXA, dual-energy X-ray absorptiometry; EPIC, Escala de

Prediccion de fracturas Implementable en historia Clínica electronica; FN, femoral neck; FRAX, Fracture Risk Assessment Tool; FRC, Fracture Risk Calculator; FREM, Fracture Risk Evaluation Model; FX, fracture; GI, gastrointestinal; MOF, major osteoporotic fracture; MORES, Male Osteoporosis Risk Estimation Score; MOST, Male Osteoporosis Screening Tool; MSCORE, Male Simple Calculated Osteoporosis Risk Estimation; NA, not applicable; NOF, National Osteoporosis Foundation instrument; OSIRIS, Osteoporosis Index of Risk; OST, Osteoporosis Self-Assessment Tool; ORAI, Osteoporosis Risk Assessment Instrument; OSTA, Osteoporosis Self-Assessment Tool for Asians; RCT, randomized clinical trial; ROB, risk of bias; ROSE, Risk-stratified Osteoporosis Strategy Evaluation; RR, relative risk; SAE, serious adverse event; SCORE, Simple Calculated Osteoporosis Risk Estimation Sn, sensitivity; SOFSURF, Study of Osteoporotic Fractures Study Utilizing Risk Factors instrument; Sp, specificity; SR, systematic review; UK, United Kingdom; VA-FARA, Veterans Affairs-Fracture Absolute Risk Assessment; WHI, Womens Health Initiative Prediction Model.

The scope of the KQ on accuracy changed between the prior report and the current update, so direct SOE comparisons are not possible. The evidence in this update was graded as low or insufficient SOE for the predictive and diagnostic accuracy of risk assessment tools and for predictive accuracy of BMD alone. Many studies were conducted using retrospectively assembled datasets of persons referred for BMD, some of whom may already have had a diagnosis of osteoporosis, been taking medication, or may have had a prior fracture. Many predictive accuracy studies focused only on discrimination outcomes and did not report sufficient information about calibration. Some used proxy data for selected risk factors or omitted those factors if data were not available, or participants were observed for fewer years than the duration used in the risk model development studies. Further, it is unclear whether data on FRAX from other countries is applicable to the US setting given that FRAX is calibrated to each country’s fracture incidence. This limitation was mitigated by restricting the KQs on predictive accuracy to countries with similar hip fracture incidence as the US. The diagnostic accuracy studies varied in how the DXA reference standard was measured (eg, different anatomic sites for BMD, different references to calculate T-scores).

Some new evidence for treatment benefits and harms was identified for this update; however, the SOE ratings for treatment benefits (KQ4) remained largely the same as the prior review: low to moderate SOE for benefit across multiple fracture outcomes for both bisphosphonates and denosumab. For treatment harms (KQ5), the SOE was low (denosumab) and moderate (bisphosphonates) for discontinuations due to AEs and serious AEs and moderate for no effect on upper gastrointestinal (GI) AEs for bisphosphonates and low for increased upper GI AEs for denosumab. As in the prior report, the SOE is insufficient for evaluating the effect of treatment on very rare harms such as osteonecrosis of the jaw, atypical femur fractures, rebound vertebral fractures, or harms of prolonged treatment duration. The major limitation in the treatment literature for primary prevention is that few studies included men, and all studies enrolled persons based on T-scores and not based on fracture risk.

A concern across the evidence for all KQs relates to the lack of diverse populations enrolled in studies. Many studies did not report the race or ethnicity of enrolled populations, and those that did mostly enrolled exclusively or majority White populations. Given the differences in fracture incidence among persons of different races and ethnicities in the United States, studies enrolling sufficient numbers from diverse populations are needed to determine the applicability of findings in different populations.

Limitations

This review focused on 1 aspect of fracture prevention: identifying and treating osteoporosis with medication. Preventing falls is addressed by a separate USPSTF recommendation.²¹⁵ This review did not address DXA testing or treatment in persons with a history of fragility fracture or medical conditions or medications associated with secondary osteoporosis. The comparative effectiveness and harms of alternative pharmacotherapies and drug holidays was not evaluated. This review was not designed to comprehensively evaluate rare harms.

CONCLUSION

Screening in higher-risk women age 65 or older was associated with a small absolute risk reduction in hip and major fractures compared with usual care. No evidence evaluated screening with BMD alone or screening in men or younger women. Risk assessment instruments, BMD alone, or both have poor to modest discrimination for predicting fracture and calibration studies were limited. Osteoporosis treatment with bisphosphonates or denosumab over several years was associated with fracture reductions and no meaningful increase in adverse events; data for longer-term or rare harms were limited based on the evidence included in this update.