If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The efficacy comparison of osteoporosis treatments can be hindered by the absence of head-to-head trials; instead, network meta-analyses (NMAs) have been used to determine comparative effectiveness. This study was the first to investigate the impact of time point–specific NMAs of osteoporosis treatments on variability in treatments’ onset of action caused by their different mechanisms of actions and trial designs.
Methods
A systematic literature review was conducted to identify randomized controlled trials (RCTs) of treatments for postmenopausal women with osteoporosis, including romosozumab (ROMO), teriparatide (TPTD), abaloparatide (ABL), alendronate (ALN), risedronate (RIS), ibandronate (IB), zoledronic acid/zoledronate (ZOL), denosumab (DEN), and raloxifene (RLX), on at least 1 fracture or bone mineral density (BMD) outcome. Of 100 RCTs identified in 5 databases, 27 RCTs were included for NMAs of new vertebral, nonvertebral, and hip fracture outcomes at 12, 24, and 36 months, and 47 RCTs were included for NMAs of BMD outcomes at lumbar spine, total hip, and femoral neck to compare the relative efficacy of osteoporosis treatments. Quality of included studies was assessed using the Cochrane Risk of Bias tool.
Findings
For vertebral fractures, TPTD (83.63%), ABL (69.11%), and ROMO/ALN (78.70%) had the highest probability to be the most effective treatment at 12, 24, and 36 months, respectively. ROMO/ALN had the highest probability (54.4%, 64.69%, and 90.29%, respectively) to be the most effective treatment for nonvertebral fractures at 12, 24, and 36 months. For hip fractures, ROMO/ALN (46.31%), ABL (61.1%), and DEN (55.21%) had the highest probability to be the most effective treatment at 12, 24, and 36 months, respectively. ROMO had the highest probability (76.06%, 44.19%, and 51.78%, respectively) to be the most effective treatment for BMD outcomes at lumbar spine, total hip, and femoral neck.
Implications
The importance of indirectly comparing available osteoporosis treatments using time point–specific NMAs was confirmed because indirect comparison results differed substantially across time points.
Osteoporosis is a progressive, systemic, skeletal disorder characterized by low bone mass, a decrease in bone strength, and increased susceptibility to fragility fractures. Osteoporosis predominantly affects postmenopausal women, with 1 in 3 women >50 years of age experiencing a fragility fracture during their lifetime compared with 1 in 5 men ≥50 years of age. The disability and mortality associated with fragility fractures represents a significant health, social, and economic burden.
Several classes of treatments are available to treat osteoporosis. Most are long-term treatments that decrease bone resorption (antiresorptive agents) and include bisphosphonates (alendronate [ALN], risedronate [RIS], ibandronate [IB], and zoledronate [ZOL]), selective estrogen receptors (raloxifene [RLX]), and antibody-based receptor activator of nuclear factor κβ (RANK) ligand inhibitor therapy (denosumab [DEN]).
Bone-building agents, such as parathyroid hormone synthetic analogs (ie, teriparatide [TPTD] and abaloparatide [ABL]), stimulate bone remodeling and increase bone mass and are used for shorter treatment periods, typically followed by an antiresorptive.
Romosozumab (ROMO) stimulates bone formation while also decreasing bone resorption. ROMO is used for shorter periods than TPTD and ABL and should be followed by an antiresorptive.
Bone-Forming and Antiresorptive Effects of Romosozumab in Postmenopausal Women With Osteoporosis: Bone Histomorphometry and Microcomputed Tomography Analysis After 2 and 12 Months of Treatment.
The efficacy of osteoporosis treatments has been assessed in multiple randomized controlled trials (RCTs), most of which compared investigational treatments with placebo (PBO). In the absence of RCTs against active comparators of interest, indirect comparative methods are commonly used to assess the comparative efficacy of different treatments. To date, several network meta-analyses (NMAs) have been conducted to determine the relative effectiveness of osteoporosis treatments.
Comparative efficacy of bone anabolic therapies in women with postmenopausal osteoporosis: A systematic review and network meta-analysis of randomized controlled trials.
Clinical effectiveness of denosumab, raloxifene, romosozumab, and teriparatide for the prevention of osteoporotic fragility fractures: A systematic review and network meta-analysis.
Indirect comparison of teriparatide, denosumab, and oral bisphosphonates for the prevention of vertebral and nonvertebral fractures in postmenopausal women with osteoporosis.
However, NMAs conducted in osteoporosis might be susceptible to considerable heterogeneity and nonsimilarity because of differing RCT populations and reporting methods (ie, robustness of NMAs in osteoporosis depends on quality of RCT data). For example, baseline characteristics, such as bone mineral density (BMD) scores, fracture history, and patients’ treatment before enrollment, may affect the comparative results of published osteoporosis NMAs.
Research has found that the relative risk (RR) of a subsequent fracture is greatest after an index fracture. Almost 50% of the subsequent fractures are expected to occur during 10 years in the first 2 years after index fracture, a period of markedly increased fracture hazard, which is referred to as the period of imminent fracture risk.
Imminent fracture risk was also revealed to be an important factor when determining the economic value of osteoporosis treatments in cost-effectiveness analyses.
With different mechanisms of action of osteoporosis treatments indicating variability in the time needed to reach meaningful fracture risk reduction, it would be expected that NMAs would be sensitive to the duration each treatment requires to reach meaningful fracture risk reductions. However, current NMA approaches in osteoporosis have often disregarded time point–specific treatment efficacy and have instead assumed equal efficacy across time points and only considered the final efficacy time point reported in each RCT. This time point can range from 12 months for bone-building treatments to 60 months for long-term bisphosphonate treatments or can even be event driven, depending on the study design.
Clinical effectiveness of denosumab, raloxifene, romosozumab, and teriparatide for the prevention of osteoporotic fragility fractures: A systematic review and network meta-analysis.
Because of their mechanisms of action, bone-building treatments, such as ROMO and TPTD, have the potential to quickly reach important fracture risk reduction benefits for patients at imminent risk of a subsequent fracture.
The impact of different time points on efficacy estimates in an NMA context has not been investigated to date. The objectives of this study were (1) to conduct a systematic literature review (SLR) of osteoporosis RCTs on fracture outcomes and changes in BMD, (2) to perform an NMA by indirectly comparing osteoporosis treatments for fracture outcomes separately at 12, 24, and 36 months and BMD outcomes, and to (3) discuss the appropriateness of performing NMAs in osteoporosis.
Materials and Methods
Systematic Literature Review
Search Strategy
An SLR was conducted to identify RCTs that reported on the efficacy of osteoporosis treatments in postmenopausal women. The search strategy adhered to guidelines published by the Cochrane Collaboration and the Centre for Reviews & Dissemination (York, United Kingdom) and are reported following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) specifications.
Medline, Medline In-Process Citations, Epubs Ahead of Print & Daily Update, EMBASE, and Cochrane Central Register of Controlled Trials (CENTRAL) were searched on September 2, 2020. Titles and abstracts were independently screened by 2 reviewers, and references that did not meet the inclusion criteria were excluded. Any discrepancy was resolved through mutual discussion or the involvement of a third reviewer. The risk of bias within each included RCT was assessed using the Cochrane Risk of Bias Tool for RCTs.
RCTs of postmenopausal women with osteoporosis were included. Treatments of interest were ROMO, TPTD, ABL, ALN, RIS, IB, ZOL, DEN, and RLX based on their licensed dose. RCTs that compared ≥2 treatments of interest or compared with PBO were included. Studies that reported at least 1 fracture (new vertebral, nonvertebral, or hip) or BMD (lumbar spine [LS], total hip [TH], or femoral neck [FN]) outcome were included. Study inclusion was not limited by language or publication date, and all publications available in the public domain were eligible for inclusion (eg, abstracts, conference posters, full reports, and journal manuscripts). Table I presents the study inclusion and exclusion criteria.
Table INMA inclusion and exclusion criteria based on PICOS.
PICOS Criteria
Inclusion Criteria
Exclusion Criteria
Population
Postmenopausal women with osteoporosis
Did not report on the population of interest
Interventions or comparators
Studies comparing at least 2 interventions of interest (plus background therapy, defined as calcium supplements and/or vitamin D): PBO, ROMO (210 mg SC every month), ROMO and ALN[ROMO/ALN] (210 mg SC every month and 70 mg every week), RLX (60 mg oral once daily), ALN (10 mg oral once daily or 70 mg oral every week), RIS (5 mg oral once daily or 35 mg oral every week), ZOL (5 mg IV yearly), DEN (60 mg SC twice yearly), TPTD (20 μg SC once daily), ABL (80 mg SC once daily), IB
Ibandronate was included only in the BMD outcomes.
(150 mg oral every month)
Did not compare at least 2 relevant interventions
Outcomes
Studies reporting appropriate data for 1 of the following outcomes: fracture outcomes at 12, 24 and 36 months, including new vertebral fracture, nonvertebral fracture, or hip fracture; BMD outcomes (percentage change at the latest time-point available from each trial): femoral neck, lumbar spine, or total hip
Did not report any relevant outcomes or did not report appropriate data (eg, RR but no 95% CrI, SD, or SE)
ABL = abaloparatide; ALN = alendronate; BMD = bone mineral density; CrI = credible interval; DEN = denosumab; IB = ibandronate; PBO = placebo; PICOS = Population, Intervention, Comparison, Outcomes and Study; RLX = raloxifene; RIS = risedronate; ROMO = romosozumab; RR = relative risk; SC = subcutanous; SD = standard deviation; SE = standard error; TPTD = teriparatide; ZOL = zoledronate.
Ibandronate was included only in the BMD outcomes.
Fracture data were reported as fracture incidence with RRs and corresponding 95% credible intervals (95% CrI) and were analyzed as dichotomous outcomes. The analyses used the incidence (ie, event data) of fractures for all fracture end points (ie, new vertebral, nonvertebral, and hip) across all treatment arms at 12, 24, and 36 months, if available. For BMD, the percentage change from baseline was reported as a mean difference (MD) with 95% CrI. Networks of evidence indicating links between osteoporosis treatments and PBO were created for every fracture end point at 12, 24, and 36 months and for BMD end points. BMD end points were not presented as time specific because of a lack of time-specific data reported in RCTs.
Missing Fracture Data
Data from the Fracture Study in Postmenopausal Women with Osteoporosis (FRAME) study
were not included in the NMA after 12 months because patients transitioned from ROMO or PBO to open-label DEN after 12 month. The ALN control arm in the Study to Determine the Efficacy and Safety of Romosozumab in the Treatment of Postmenopausal Women With Osteoporosis (ARCH)
allowed ROMO to be connected with the network of evidence after 12 months under the following assumptions: the incorporation of the Fosamax International Trial data
Multinational, placebo-controlled, randomized trial of the effects of alendronate on bone density and fracture risk in postmenopausal women with low bone mass: results of the FOSIT study. Fosamax International Trial Study Group.
based on a median of 33 months were included at 36 months for ROMO/ALN on nonvertebral and hip fractures. Other assumptions include the inclusion of the Abaloparatide Comparator Trial In Vertebral Endpoints (ACTIVE)
(duration of 10.3 months) at 12 months for hip fractures; no data for TPTD or ABL were available at 36 months. As suggested by National Institute for Health and Care Excellence (NICE), zero correction was applied when zero events were reported in a treatment arm.
NICE Decision Support Unit Technical Support Documents. A Generalised Linear Modelling Framework for Pairwise and Network Meta-Analysis of Randomised Controlled Trials.
National Institute for Health and Care Excellence (NICE),
London2014
Statistical heterogeneity in direct head-to-head meta-analyses was assessed using the I2 test. The I2 test describes the percentage of total variance across RCTs attributable to heterogeneity rather than chance.
The similarity assumption was evaluated by comparing differences in RCT baseline characteristics (ie, mean age, proportion of individuals with prevalent fracture, and mean BMD), treatment dosages and durations, comparators, and end point definitions. The consistency was assessed based on the NICE Decision Support Unit Technical Support Document 4.
On the basis of NICE Decision Support Unit Technical Support Document 2, all analyses were conducted within a bayesian framework using the statistical software BUGS (WinBUGS).
NICE Decision Support Unit Technical Support Documents. A Generalised Linear Modelling Framework for Pairwise and Network Meta-Analysis of Randomised Controlled Trials.
National Institute for Health and Care Excellence (NICE),
London2014
Binary bayesian models were used for fracture end points, and shared parameter bayesian models were used for BMD end points. NMAs were conducted using both fixed-effects and random-effects models. The statistical model fit was determined by evaluating deviance information criteria values.
NICE Decision Support Unit Technical Support Documents. A Generalised Linear Modelling Framework for Pairwise and Network Meta-Analysis of Randomised Controlled Trials.
National Institute for Health and Care Excellence (NICE),
London2014
NICE Decision Support Unit Technical Support Documents. A Generalised Linear Modelling Framework for Pairwise and Network Meta-Analysis of Randomised Controlled Trials.
National Institute for Health and Care Excellence (NICE),
London2014
All analyses were performed with 50,000 iterations after a burn-in of 30,000 Markov Chain Monte Carlo simulations, using 2 chains. NMA diagnostics (ie, trace, quantiles, auto cor) were checked to ensure sufficiently converged data; otherwise, an additional 50,000 iterations were performed.
NICE Decision Support Unit Technical Support Documents. A Generalised Linear Modelling Framework for Pairwise and Network Meta-Analysis of Randomised Controlled Trials.
National Institute for Health and Care Excellence (NICE),
London2014
Treatment ranking probabilities were estimated based on the size of treatment effect, using the number of iterations where a treatment has a specific rank over the total number of iterations of the Markov chain.
The bayesian approach in WinBUGS captures the uncertainty around the treatment effects by running multiple iterations, where the size of treatment effect and thus the rank changes with each iteration due to the given variance (vague priors).
Forest plots, heterogeneity I2 tests, and consistency tests were conducted using an integrated development environment for R, (version 4.0.2) (RStudio, Boston, Massachusetts).
Pairwise comparisons reporting RR point estimates with corresponding 95% CrIs of vertebral, nonvertebral, and hip fracture risk reduction at 12, 24, and 36 months were presented in league tables. Pairwise comparisons in league tables reporting MDs with corresponding 95% CrI of change in BMD at LS, TH, and FN were also presented. The networks of evidence, forest plots, and ranking tables, including probabilities of each treatment being the most effective, were reported for all fracture end points at all time points and for all BMD end points.
Results
Literature Search
A total of 30,345 citations were retrieved through searching electronic databases, conference abstracts, trial registries, and other sources, including gray literature. An additional 8 records were retrieved through hand-searching. After duplicate removal (8421 duplicate studies), 21,932 citations were independently screened by 2 reviewers. From these, 259 citations reporting on a total of 100 RCTs were judged as meeting the inclusion criteria. A summary of the study inclusion process is presented in the Figure in accordance with PRISMA guidelines.
Treatments reported by the RCTs on fracture data included ABL, ALN, DEN, RLX, RIS, ROMO, TPTD, and ZOL. Treatments reported by RCTs on BMD data included ABL, ALN, DEN, IB, RLX, RIS, ROMO, TPTD, and ZOL. A summary of all RCTs on fracture and BMD data is reported in Supplemental Tables II and III.
Overview of Trial and Population Characteristics
Most of the RCTs were conducted in multicentered sites around the world and were industry sponsored. Most of the included RCTs were randomized, Phase III, double-blinded trials with a follow-up period of 12 to 60 months. Detailed information on RCT characteristics for fracture and BMD RCTs are presented in Supplemental Table IV. The RCTs varied with respect to their inclusion criteria for participants, with the exception that they all sought to recruit postmenopausal women at risk of osteoporotic fracture. The most frequently described inclusion criteria focused on restricting RCT populations in terms of age, number of years since menopause, fracture history, and range of BMD or T scores (ie, scores that determine bone health in patients by comparing bone mass of patients at risk with bone mass in healthy young people, where osteoporosis is reflected by a T score of <–2.5).
An overview of the population characteristics for fracture and BMD outcomes is presented in Supplemental Tables V and VI, respectively.
Trial Quality Assessment
The risk of bias of the studies included in the analyses of fracture and BMD outcomes was assessed using the Cochrane risk of bias tool.19 RCT quality varied substantially, with only 3 RCTs for fracture outcomes and 2 RCTs for BMD outcomes being assessed as having an overall low risk of bias (ARCH,
Effect of alendronate on risk of fracture in women with low bone density but without vertebral fractures: results from the Fracture Intervention Trial.
). Eighteen and 34 RCTs were judged to be at high risk of bias for ≥1 domains for fracture and BMD outcomes, respectively, with the remaining RCTs assessed at low or unclear risk of bias. Supplemental Table VII presents the risk of bias for each RCT included in the analyses.
Network Meta-analyses
Analyses were performed in cases in which suitable fracture or BMD data were reported by included RCTs. A total of 27 RCTs that reported fractures and 42 RCTs that reported BMD were therefore included in the analyses. Other identified fracture and BMD trials were not included in the analyses because these were mainly extension studies or studies on subgroups (eg, ethnic subgroups). Reported outcomes in the included RCTs consisted of vertebral fractures in 20 of 27 (57.1%), nonvertebral fractures in 15 of 27 (42.9%), hip fractures in 15 of 27 (42.9%), LS BMD in 41 of 42 (97.6%), TH BMD in 30 of 42 (71.4%), and FN BMD in 31 of 42 (73.8%). The studies included in each of the analyses for the fracture and BMD outcomes are presented in Supplemental Tables VIII and IX, respectively.
Network diagrams for fracture outcomes at each time point and BMD outcomes are presented in Supplemental Figures 1 and 2, respectively. Heterogeneity and inconsistency analyses for the fracture outcomes are presented in Supplemental Tables 10 and 11. The fixed-effects model estimates are presented for all fracture outcomes after evaluating model fit using deviance information criteria. Because of the significant amount of heterogeneity among the studies used in the BMD analyses, the random-effects models estimates are reported.
Fracture Outcomes
A summary of fracture outcome results is given in Table II. The treatments with the highest and second highest probability of being the most effective treatment by fracture outcome and time point are reported. More detailed outcomes relating to each fracture type are given below.
Table IISummary of treatments with the highest and second-highest probability of being the most effective treatment by fracture outcome and time point.
Fracture Outcome Time Point, months
Vertebral Fractures
Nonvertebral Fractures
Hip Fractures
12
TPTD and ROMO
ROMO and TPTD
ROMO and DEN
24
ABL and ROMO/ALN
ROMO/ALN and ABL
ABL and RIS
36
ROMO/ALN and ZOL
ROMO/ALN and RIS
DEN and ROMO/ALN
ABL = abaloparatide; ALN = alendronate; DEN = denosumab; RIS = risedronate; ROMO = romosozumab; ROMO/ALN = romosozumab followed by alendronate; TPTD = teriparatide; ZOL = zoledronate.
At 12 months, 7 of 7 active treatments were associated with a statistically significant reduction in new vertebral fracture risk versus PBO. TPTD and ROMO were the 2 treatments with the greatest reduction in new vertebral fracture risk versus PBO (RR = 0.20; 95% CrI, 0.09–0.38; RR = 0.31; 95% CrI, 0.18–0.48, respectively). Of 8 treatments, TPTD had the highest (83.63%) and ROMO the second-highest probability of being the most effective treatment (13.22%).
At 24 months, 8 of 8 active treatments were associated with a statistically significant reduction in new vertebral fracture risk versus PBO. ABL and ROMO/ALN were the 2 treatments with the greatest reduction in new vertebral fracture risk versus PBO (RR = 0.19; 95% CrI, 0.04–0.34; RR = 0.19; 95% CrI, 0.12–0.28, respectively). Of 9 treatments, ABL had the highest (69.11%) and ROMO/ALN the second-highest probability of being the most effective treatment (27.68%).
At 36 months, 6 of 6 active treatments were associated with a statistically significant reduction in new vertebral fracture risk versus PBO. ROMO/ALN and ZOL were the treatments with the greatest reduction in new vertebral fracture risk versus PBO (RR = 0.24; 95% CrI, 0.16–0.35; RR = 0.30; 95% CrI, 0.23–0.37, respectively). Of 7 treatments, ROMO/ALN had the highest (78.70%) and ZOL the second-highest probability of being the most effective treatment (16.54%).
The league tables, including all pairwise RR comparisons for new vertebral fractures, are presented in Table III. Supplemental Figure 3 presents the forest plots depicting RRs and the median ranking of each treatment versus PBO at 12, 24, and 36 months. The ranking probabilities for each treatment are presented in Supplemental Table XII.
Table IIIPairwise comparisons (relative risk [95% credible interval]) for new vertebral fractures at 12, 24, and 36 months.
At 12 months, only 1 of 7 active treatments (ROMO) was associated with a statistically significant reduction in nonvertebral fracture risk versus PBO (RR = 0.64; 95% CrI, 0.47–0.86). TPTD, although not statistically significant, had the second greatest reduction in nonvertebral fracture risk versus PBO (RR = 0.75; 95% CrI, 0.45–1.16). Of 8 treatments, ROMO had the highest (54.40%) and TPTD the second-highest probability of being the most effective treatment (26.16%).
At 24 months, 6 of 8 active treatments were associated with a statistically significant reduction in nonvertebral fracture risk versus PBO. ROMO/ALN and ABL were the treatments with the greatest reduction in nonvertebral fracture risk versus PBO (RR = 0.43; 95% CrI, 0.22–0.74; RR = 0.52; 95% CrI, 0.29–0.84, respectively). Of 9 treatments, ROMO/ALN had the highest (64.96%) and ABL the second-highest probability to be the most effective treatment (29.38%).
At 36 months, 5 of 6 active treatments were associated with a statistically significant reduction in nonvertebral fracture risk versus PBO. ROMO/ALN and ALN were the treatments with the greatest reduction in nonvertebral fracture risk versus PBO (RR = 0.51; 95% CrI, 0.39–0.65; RR = 0.62; 95% CrI, 0.54–0.72, respectively). Of 7 treatments, ROMO/ALN had the highest (90.29%) and RIS the second-highest probability to be the most effective treatment (8.19%).
The league tables of all pairwise comparisons for nonvertebral fractures are presented in Table IV. Supplemental Figure 4 presents the forest plots depicting RRs and the median ranking of each treatment versus PBO at 12, 24, and 36. The ranking probabilities for each treatment are presented in Supplemental Table XII.
Table IVPairwise comparisons (relative risks [95% credible intervals]) for nonvertebral fractures at 12, 24, and 36 months.
At 12 months, none of the 5 active treatments were associated with a statistically significant reduction in hip fracture risk versus PBO. Although neither were statistically significant, ROMO and DEN were the treatments with the greatest reduction in hip fracture risk versus PBO (RR = 0.54; 95% CrI, 0.19–1.16; RR = 0.55; 95% CrI, 0.25–1.02, respectively). Of 6 treatments, ROMO had the highest (46.31%) and DEN the second-highest probability of being the most effective treatment (42.04%).
At 24 months, 4 of 8 active treatments were associated with statistically significant reduction in hip fracture risk versus PBO. ABL and ROMO/ALN were the treatments with the greatest reduction in hip fracture risk versus PBO (RR = 0.36; 95% CrI, 0.01–2.18; RR = 0.37; 95% CrI, 0.14–0.79, respectively). Of 9 treatments, ABL had the highest (61.10%) and RIS the second-highest probability of being the most effective treatment (13.34%), which is appraised in more detail in the discussion.
At 36 months, 4 of 6 active treatments were associated with statistically significant reduction in hip fracture risk versus PBO. DEN and ROMO/ALN were the treatments with the greatest reduction in hip fracture risk versus PBO (RR = 0.28; 95% CrI, 0.18–0.43; RR = 0.32; 95% CrI, 0.12–0.67, respectively). Of 7 treatments, DEN had the highest (55.21%) and ROMO/ALN the second-highest probability of being the most effective treatment (44.63%).
The league tables of all pairwise comparisons for hip fractures are presented in Table V. Supplemental Figure 5 presents the forest plots depicting RRs and the median ranking of each treatment versus PBO at 12, 24, and 36 months. The ranking probabilities for each treatment are presented in Supplemental Table XII.
Table VPairwise comparisons (relative risks [95% credible intervals]) for hip fractures at 12, 24, and 36 months.
BMD outcomes were not time point specific, owing to paucity of data. In the analysis of LS BMD, 9 of 9 active treatments were associated with a statistically significant change in LS BMD versus PBO. ROMO and ZOL were the treatments with the greatest change in LS BMD versus PBO (MD = 11.86; 95% CrI, 8.18–15.59; MD = 8.98; 95% CrI, 5.88–12.13, respectively). Of 10 treatments, ROMO had the highest (76.06%) and ABL the second-highest probability of being the most effective treatment (13.28%).
In the analysis of TH BMD, 6 of 9 active treatments were associated with a statistically significant change in TH BMD versus PBO. ROMO and ZOL were the treatments with the greatest change in TH BMD versus PBO (MD = 5.76; 95% CrI, 3.78–7.76; MD = 5.74; 95% CrI, 3.93–7.67, respectively). Of 10 treatments, ROMO had the highest (ie, 44.19%) and ZOL the second-highest probability to be the most effective treatment (ie, 41.28%).
In the analysis of FN BMD, 9 of 9 active treatments were associated with a statistically significant change in FN BMD versus PBO. ROMO and ZOL were the treatments with the greatest change in FN BMD versus PBO (MD = 5.63; 95% CrI, 4.28–6.97; MD = 5.58; 95% CrI, 4.04–7.24, respectively). Of 10 treatments, ROMO had the highest (51.78%) and ZOL the second-highest probability to be the most effective treatment (46.27%).
The league tables of all pairwise MD comparisons on BMD change are presented in Table VI. Supplemental Figure 6 presents the MD in change in BMD and the median ranking of each treatment versus PBO. The ranking probabilities for each treatment are presented in Supplemental Table XII.
Table VIPairwise comparisons (mean differences [95% credible intervals]) of bone mineral density outcomes.
This study aimed to investigate the impact of different time points ofefficacy estimates in an NMA. To do so, we conducted a rigorous and broad SLR in which searches were not limited by date of publication, language, or outcomes and which adhered to accepted guidelines to reduce the risk of bias. Our SLR identified 35 fracture and 65 nonfracture RCTs, of which 27 and 42 RCTs were included in our NMAs on fracture and BMD outcomes, respectively. The NMAs were conducted using published and validated WingBUGS codes according to NICE technical guidelines (NICE Decision Support Unit Technical Support Documents 2 and 4).
Our NMA found that at different time points, ranging from 12 to 36 months, different treatments had the highest probability of being the most effective treatment for fracture outcomes, confirming the importance of time point when making these assessments. The NMA revealed TPTD, ABL, and ROMO/ALN to have highest probability of being the most effective treatment for vertebral fractures at 12, 24, and 36 months, respectively. For nonvertebral fractures, ROMO/ALN had the highest probability of being the most effective treatment across all time points. For hip fractures, ROMO, ABL, and DEN had the highest probability of being the most effective treatment at 12, 24, and 36 months, respectively. Across time points and fracture sites, ROMO/ALN was the treatment most frequently reported as having either the highest or second-highest probability of being the most effective treatment. For BMD outcomes, which are strongly associated with fracture reductions, the treatment with the greatest number of statistically significant pairwise comparisons was ROMO, with statistically significant changes in BMD versus 6 of 9 treatment comparators across every BMD site.
Relationship Between Bone Mineral Density and Risk of Vertebral Fractures with Denosumab Treatment in Japanese Postmenopausal Women and Men with Osteoporosis.
The NMA did not reveal statistically significant differences in reduction of fracture risk between bone-building treatments (ROMO, TPTD, and ABL). However, pairwise comparisons found that ROMO/ALN was more frequently statistically significant versus non–bone-building treatments than TPTD and ABL; namely, ROMO/ALN had a statistically significant fracture risk reduction in the pairwise comparisons at 24 months versus ZOL for new vertebral fractures, versus DEN and ZOL for nonvertebral fractures, and versus PBO for hip fractures, which TPTD and ABL did not have. There was only 1 scenario in which TPTD had a statistically significant fracture risk reduction against a non–bone-building treatment, and ROMO/ALN did not (vs RIS for hip fractures at 12 months). Although ROMO/ALN achieved statistically significant fracture risk reduction at 36 months versus PBO, RLX, ALN, and RIS for new vertebral fractures, versus PBO, RLX, ALN, ZOL, and DEN for nonvertebral fractures, and versus PBO, RLX, and ALN for hip fractures, there was no fracture reduction evidence from RCTs at 36 months for TPTD or ABL. ROMO had statistically significant improvements in BMD over TPTD across all BMD sites (LS, TH, and FN).
The robustness of any NMA depends on the quality of RCT data available. NMAs in osteoporosis are hampered by a number of factors, including (1) the absence of consistent end point reporting standards in RCTs (ie, differing fracture-type definitions [some RCTs recorded fracture outcomes as the number of patients with a fracture and others as the time to fracture] and varying time points of assessments); (2) differences in reporting durations (because of different MoAs: bone-building treatments with shorter periods and long-term treatments, such as bisphosphonates, DEN, and RLX, with longer treatment periods); (3) lack of power in some RCTs to detect meaningful differences in hip fracture incidences (eg, in ACTIVE
Effects of teriparatide and risedronate on new fractures in post-menopausal women with severe osteoporosis (VERO): a multicentre, double-blind, double-dummy, randomised controlled trial.
which results in wide CrIs in specific estimates in the NMAs); and (4) intertrial differences in baseline characteristics (eg, fracture history, BMD, age, and treatment history). Therefore, comparative benefits observed in NMAs may not only be caused by differences in the mechanisms of action of the treatments but also attributable to differing trial designs. Conducting time-sensitive NMAs in osteoporosis can help to address this issue, but future research is needed to investigate the precise impact of differing trial designs on NMAs.
Despite following published methodologic guidelines, this study was associated with some limitations. First, the risk of bias assessment found that reporting of randomization and allocation concealment methods was particularly poor across included RCTs. Concealment methods are important when fracture outcomes are based on the assessment and interpretation of radiographic findings because this might otherwise introduce bias. The impact of RCTs with a high risk of bias on the NMA results was not assessed. Second, there was evidence of moderate heterogeneity in 3 fracture end points (ie, new vertebral fractures at 12 months [RLX vs PBO], nonvertebral fractures at 12 months [ZOL vs PBO], and hip fractures at 24 months [RIS vs TPTD]). However, no evidence of inconsistency was observed across included RCTs. The BMD networks were larger and more complex, and the heterogeneity and inconsistency among RCTs might have influenced the results, warranting careful interpretation. Third, the validity of the NMA might have been affected by the end point reporting standards in RCTs and the quantity and quality of the included data available from the published RCTs; in particular, the data scarcity for hip fractures, because of RCTs in osteoporosis not being designed or powered to detect meaningful hip fracture reduction levels (eg, TPTD and ABL), led to wide CrIs, indicating uncertainty of comparative results. The very favorable NMA outcomes of RIS for hip fractures at 24 months are questionable because they are not consistent with RCT data and are at risk of being artificially inflated because of the very low incidence of hip fractures observed in the VERO trial (0.73%) and the trial by Hadji et al, (0.57%) combined with the inconsistent and directionally opposite hazard ratios of RIS vs TPTD (2.50 and 0.41 in the VERO trial and the trial by Hadji et al, respectively).
Effects of teriparatide and risedronate on new fractures in post-menopausal women with severe osteoporosis (VERO): a multicentre, double-blind, double-dummy, randomised controlled trial.
Therefore, outcomes from earlier time points with larger CrIs because of a low number of events may result in higher uncertainty and need to be interpreted with caution.
A comparison with previously published NMAs to determine the validity of this study was difficult because this is the first study to conduct time-specific NMAs; however, previous findings that bone-building treatments most frequently have the greatest comparative fracture risk reduction and improvements in BMD were confirmed in this study. Simpson et al
Clinical effectiveness of denosumab, raloxifene, romosozumab, and teriparatide for the prevention of osteoporotic fragility fractures: A systematic review and network meta-analysis.
estimated the RRs for ROMO/ALN vs PBO at 0.25 (95% CrI, 0.15–0.43) for vertebral fractures, 0.71 (95% CrI, 0.48–0.85) for nonvertebral fractures, and 0.39 (95% CrI, 0.21–0.72) for hip fractures, which is comparable to the estimate of this NMA at 24 months: 0.19 (95% CrI, 0.12–0.28) for vertebral fractures, 0.43 (95% CrI, 0.22–0.74) for nonvertebral fractures, and 0.37 (95% CrI, 0.14–0.79) for hip fractures for ROMO/ALN versus PBO. None of the previously published NMAs included all treatments and time-specific endpoints as was done in this study. Most published NMAs focused on the indirect comparison of bisphosphonates (ie, ALN, RIS, IB, and ZOL) and did not include bone-building agents, such as ROMO, TPTD and ABL. Overall, these studies found that bisphosphonates, such as ZOL and ALN, were most effective in reducing the risk of fractures and increasing BMD compared with PBO. Our study similarly indicated that ZOL was an important bisphosphonate in increasing BMD at all BMD sites.
Comparative efficacy of bone anabolic therapies in women with postmenopausal osteoporosis: A systematic review and network meta-analysis of randomized controlled trials.
focused on the comparative efficacy of ROMO, TPTD, and ABL with PBO and bisphosphonates (ie, RIS, ALN, and ZOL were pooled in this study). This study confirmed that that bone-building agents were the most effective treatments in fracture risk reductions across most fracture sites (ie, vertebral and nonvertebral fractures). ROMO, TPTD, and ABL significantly reduced the risk on vertebral fractures, whereas in contrast to our study, none of the bone-building agents significantly reduced nonvertebral fractures because of wider 95% CrIs of the RRs. Bone-building agents significantly increased BMD at all sites (ie, LS, TH, and FN) compared with PBO and ROMO ranked better at all BMD locations, which is in line with the presented BMD results.
Comparative efficacy of bone anabolic therapies in women with postmenopausal osteoporosis: A systematic review and network meta-analysis of randomized controlled trials.
Because quality varies across RCTs in osteoporosis, an additional comparative analysis with 10 similar previously published SLR/NMA studies on adjusting for baseline characteristics was conducted.
Comparative efficacy of bone anabolic therapies in women with postmenopausal osteoporosis: A systematic review and network meta-analysis of randomized controlled trials.
Clinical effectiveness of denosumab, raloxifene, romosozumab, and teriparatide for the prevention of osteoporotic fragility fractures: A systematic review and network meta-analysis.
A network meta-analysis on the short-term efficacy and adverse events of different anti-osteoporosis drugs for the treatment of postmenopausal osteoporosis.
Clinical effectiveness of denosumab, raloxifene, romosozumab, and teriparatide for the prevention of osteoporotic fragility fractures: A systematic review and network meta-analysis.
A network meta-analysis on the short-term efficacy and adverse events of different anti-osteoporosis drugs for the treatment of postmenopausal osteoporosis.
performed subgroup analyses to find any effect of variables that might affect the treatment effect. However, none of the studies detected a significant impact of potential effect modifiers, such as age and sex. Subgroup analyses found that overall findings were robust and were not affected by exclusion of any study that was contributing the most heterogeneity. Some of these studies performed subgroup analyses stratified by treatment duration or type of bisphosphonate; however, this research presented time-specific NMAs. Because previous studies included similar RCTs, the current approach of using time-specific NMAs without replication of meta-regressions or subgroup analyses was considered robust.
Most RCTs in osteoporosis were designed to examine the benefits of long-term treatments rather than rapid responses to short-term treatments, such as bone-building agents (eg, TPTD, ABL, and ROMO).
Time-specific NMAs providing comparative efficacy results at time points as early as 12 months may provide important information for treatment decision making for patients who incurred a recent fracture, are at high imminent fracture risk, and could benefit the most from treatments with fast-acting benefits. The NMA results of this study revealed the treatment rankings and pairwise comparisons to differ across time points, with various treatments reaching levels of statistical significance at different time points. These findings highlight the importance of indirectly comparing available osteoporosis treatments using time-specific NMAs.
However, it remains important to differentiate between long- and short-term treatments when conducting time-specific NMAs because the efficacy improves over time with long-term treatments. In addition, the ranking of treatments is bound to RCT data availability because not all treatments were evaluated at all time points for each end point. Therefore, ranking of treatments should not be considered without the pairwise comparisons when interpreting time-specific NMA results for osteoporosis treatments. If NMAs do not account for different follow-up periods of osteoporosis RCTs, they may be at risk of underestimating or misstating the comparative benefits of the included treatments. This in turn can also lead to the economic value of osteoporosis treatments being underestimated in cost-effectiveness analyses in which NMAs are recommended to be used as a source of efficacy in the absence of head-to-head trials.
Recommendations for the conduct of economic evaluations in osteoporosis: outcomes of an experts' consensus meeting organized by the European Society for Clinical and Economic Aspects of Osteoporosis, Osteoarthritis and Musculoskeletal Diseases (ESCEO) and the US branch of the International Osteoporosis Foundation.
Efficacy outcomes reported at different time points in osteoporosis RCTs should be considered when performing indirect treatment comparisons because our study found that the outcomes of NMAs can vary by time point. Time-specific NMAs allow for comparison of outcomes at different time points, which is relevant in treatment decision-making. Bone-building treatments, such as ROMO, TPTD, and ABL, were the most effective treatments in fracture risk reductions across most fracture sites and time points, with ROMO/ALN being the treatment to most frequently have the highest or second-highest probability of being the most effective across all fracture sites and time points. Considerations should be given to the methodologic limitations of published RCTs when conducting and interpreting indirect comparative analyses.
Acknowledgments
Author contributions are as follows: substantial contributions to study conception and design: D. Willems, M.K. Javaid, R. Pinedo-Villaneuva, C. Libanati, A. Yehoshua, M. Charokopou; substantial contributions to analysis and interpretation of the data: D. Willems, M.K. Javaid, R. Pinedo-Villaneuva, C. Libanati, A. Yehoshua, M. Charokopou; drafting the article or revising it critically for important intellectual content: D. Willems, M.K. Javaid, R. Pinedo-Villaneuva, C. Libanati, A. Yehoshua, M. Charokopou; final approval of the version of the article to be published: D. Willems, M.K. Javaid, R. Pinedo-Villaneuva, C. Libanati, A. Yehoshua, M. Charokopou. H. El Alili, F. Kroi, S. van Beekhuizen (Ingress-Health) additionally supported the analyses and writing of the manuscript.
Funding Sources
This study was funded by UCB Pharma and Amgen Inc.
Disclosure
D. Willems is employed by UCB Pharma; M.K. Javaid received honoraria, unrestricted research grants, and travel and/or subsistence expenses from AbbVie, Amgen, Consilient Health, Kyowa Kirin, and UCB Pharma; R. Pinedo-Villaneuva received lecture fees and/or consulting honoraria from Amgen, Kyowa Kirin Services, Mereo Biopharma, and UCB Pharma; C. Libanati and M. Charokopou are employed by and stockholders of UCB Pharma; A. Yehoshua is employed by Amgen Inc. H. El Alili, F. Kroi, S. van Beekhuizen are full time employees of Ingress-Health, who received consulting fees from UCB Pharma and Amgen. The authors have indicated that they have no other conflicts of interest.
Bone-Forming and Antiresorptive Effects of Romosozumab in Postmenopausal Women With Osteoporosis: Bone Histomorphometry and Microcomputed Tomography Analysis After 2 and 12 Months of Treatment.
Comparative efficacy of bone anabolic therapies in women with postmenopausal osteoporosis: A systematic review and network meta-analysis of randomized controlled trials.
Clinical effectiveness of denosumab, raloxifene, romosozumab, and teriparatide for the prevention of osteoporotic fragility fractures: A systematic review and network meta-analysis.
Indirect comparison of teriparatide, denosumab, and oral bisphosphonates for the prevention of vertebral and nonvertebral fractures in postmenopausal women with osteoporosis.
Multinational, placebo-controlled, randomized trial of the effects of alendronate on bone density and fracture risk in postmenopausal women with low bone mass: results of the FOSIT study. Fosamax International Trial Study Group.
NICE Decision Support Unit Technical Support Documents. A Generalised Linear Modelling Framework for Pairwise and Network Meta-Analysis of Randomised Controlled Trials.
National Institute for Health and Care Excellence (NICE),
London2014
Effect of alendronate on risk of fracture in women with low bone density but without vertebral fractures: results from the Fracture Intervention Trial.
Relationship Between Bone Mineral Density and Risk of Vertebral Fractures with Denosumab Treatment in Japanese Postmenopausal Women and Men with Osteoporosis.
Effects of teriparatide and risedronate on new fractures in post-menopausal women with severe osteoporosis (VERO): a multicentre, double-blind, double-dummy, randomised controlled trial.
A network meta-analysis on the short-term efficacy and adverse events of different anti-osteoporosis drugs for the treatment of postmenopausal osteoporosis.
Recommendations for the conduct of economic evaluations in osteoporosis: outcomes of an experts' consensus meeting organized by the European Society for Clinical and Economic Aspects of Osteoporosis, Osteoarthritis and Musculoskeletal Diseases (ESCEO) and the US branch of the International Osteoporosis Foundation.
Ibandronate was included only in the BMD outcomes.
00489-6&id=gr1.jpg){kind=link}