Why did FREEDOM use placebo instead of alendronate as the comparator?

At the time of trial design (2004), FDA guidance accepted placebo-controlled fracture endpoints for anti-resorptive agents. An active-comparator trial would have required roughly double the sample size to show a difference. The result is that FREEDOM proves denosumab reduces fractures versus no treatment, not that it outperforms bisphosphonates.

What does 'morphometric vertebral fracture' mean in FREEDOM?

It refers to a vertebral fracture identified on scheduled lateral spine X-rays using the Genant semiquantitative method, which requires at least a 20% reduction in vertebral body height. This includes fractures the patient may not have felt.

How were fractures assessed to prevent bias?

All lateral spine radiographs were read at a single central facility by readers blinded to treatment assignment. This eliminated variability between study sites and reduced the chance of detection bias influencing the primary endpoint.

What was the absolute risk reduction in vertebral fractures?

The 3-year vertebral fracture rate was 2.3% with denosumab versus 7.2% with placebo, giving an absolute risk reduction of 4.9 percentage points and a number needed to treat of approximately 20 over three years.

Did FREEDOM detect osteonecrosis of the jaw or atypical femoral fractures?

No. Both events were too rare to appear in a 3-year, 7,868-patient trial. ONJ cases were later reported in the open-label FREEDOM Extension, with a cumulative incidence of about 5.2 per 10,000 patient-years over up to 10 years of treatment.

What happened to patients who stopped denosumab after FREEDOM?

FREEDOM itself did not study discontinuation effects. Later analyses from the extension study and post-marketing data revealed that stopping denosumab causes rapid bone density loss and a rebound increase in vertebral fractures, leading to current recommendations to transition to a bisphosphonate upon cessation.

Does FREEDOM apply to men with osteoporosis?

Not directly. FREEDOM enrolled exclusively postmenopausal women. Denosumab's approval for male osteoporosis rests on the separate ADAMO trial, which used BMD as the primary endpoint rather than fracture reduction.

How did FREEDOM handle missing data and dropouts?

The primary analysis used a modified intention-to-treat population: all randomized patients who received at least one dose and had baseline plus at least one post-baseline vertebral X-ray. About 24% of patients did not complete the 36-month visit. The impact of these missing data on the primary result is likely small but introduces some uncertainty.

Was the 68% reduction clinically meaningful or just statistically significant?

Both. The absolute risk reduction of 4.9% over 3 years (NNT of ~20) represents a meaningful clinical benefit. The hierarchical testing procedure also confirmed significant reductions in hip fracture (40%) and nonvertebral fracture (20%), reinforcing that the skeletal benefit extends beyond the spine.

How does FREEDOM compare to bisphosphonate fracture trials in design quality?

FREEDOM's design is comparable to landmark bisphosphonate trials like FIT (alendronate) and HORIZON (zoledronic acid) in sample size, duration, and endpoint rigor. All three used central radiographic reading and morphometric vertebral fracture endpoints. The main design difference is that HORIZON included a clinical fracture co-primary endpoint, while FREEDOM relied on hierarchical secondary testing for nonvertebral outcomes.

Inside the FREEDOM Methodology: What Most Summaries Skip

By HealthRX Medical Team

Published May 25, 2026Updated May 25, 2026Last reviewed May 25, 2026

At a glance

| Detail | Value | |---|---| | N | 7,868 | | Intervention | Denosumab 60 mg subcutaneous every 6 months | | Comparator | Placebo (both groups received calcium + vitamin D) | | Duration | 36 months | | Primary endpoint | New morphometric vertebral fractures at 36 months | | Key result | 68% relative risk reduction (2.3% vs 7.2%; p <0.001) | | Registration | NCT00089791 |

Why the Abstract Is Not Enough

Most secondary sources repeat two numbers from the FREEDOM publication: 68% vertebral fracture reduction and 7,868 patients. That framing hides at least six design decisions that determine what the trial actually proved, who it applies to, and where its evidence stops. This page unpacks each one.

Population Selection: Who Got In and Who Didn't

FREEDOM enrolled postmenopausal women aged 60 to 90 with a lumbar spine or total hip T-score between −2.5 and −4.0 at either site. That lower bound matters. Women with T-scores below −4.0 were excluded, removing the most severely osteoporotic patients, the group at highest absolute fracture risk.

The upper bound of −2.5 excluded osteopenia. The trial population therefore represents a defined band of moderate-to-severe osteoporosis, not the full clinical spectrum. Exclusion criteria also removed women who had used oral bisphosphonates for more than 3 years, IV bisphosphonates within 12 months, or any prior use of strontium ranelate, PTH, or fluoride. This created a treatment-naive or minimally-treated cohort, ideal for isolating denosumab's effect but not reflective of real-world sequencing where patients often switch from a bisphosphonate.

Calcium (≥1 to 000 mg) and vitamin D (≥400 IU) supplementation was required in both arms. This background therapy raises the effective comparator from pure placebo to "placebo plus supplementation," an important distinction when comparing FREEDOM results to trials where supplementation was not standardized.

Randomization and Blinding

The trial used a 1:1 randomization stratified by study site. Block sizes were not publicly reported in the primary manuscript, though the protocol registration confirms central randomization via an interactive voice-response system. Both patients and investigators were blinded; the prefilled syringes for denosumab and placebo were visually identical.

One subtlety: denosumab suppresses the bone turnover marker CTX rapidly and dramatically. Any investigator who ordered bone turnover markers could have been functionally unblinded. The protocol did not prohibit ordering these labs, though centralized assessment of the primary endpoint (vertebral morphometry by a blinded radiology center) mitigated the impact of any site-level unblinding on the primary outcome.

The Primary Endpoint: Morphometric Vertebral Fractures

FREEDOM defined its primary endpoint as new morphometric vertebral fractures assessed by lateral spine radiographs at baseline, 12, 24, and 36 months. Fractures were graded using the semiquantitative Genant method, where a new fracture requires ≥20% reduction in vertebral height from baseline.

This choice has three implications that most summaries ignore:

The HealthRX Endpoint Interpretation Framework for FREEDOM:

| Design choice | What it captures | What it misses | |---|---|---| | Morphometric (radiographic) endpoint | Subclinical and clinical vertebral fractures alike | Symptom burden; many morphometric fractures are painless | | Genant semiquantitative grading | Standardized, reproducible across centers | Borderline (<20% height loss) fractures excluded; threshold is somewhat arbitrary | | Scheduled X-rays at fixed intervals | Uniform ascertainment, reducing detection bias | Does not capture exact fracture timing; misses fractures between visits if vertebra remodels | | Central reading facility | Eliminates inter-site reader variability | Adds cost and logistic complexity |

The 68% relative risk reduction (2.3% denosumab vs 7.2% placebo) therefore describes radiographic vertebral fractures captured at predetermined intervals by a central reader using the Genant cutoff. Clinical vertebral fractures, the subset that actually cause symptoms, were a secondary endpoint and showed a 69% reduction, a reassuring consistency. The absolute risk reduction of 4.9 percentage points over 3 years translates to a number needed to treat (NNT) of approximately 20 to prevent one morphometric vertebral fracture.

Secondary Endpoints and the Multiplicity Problem

FREEDOM pre-specified a hierarchical testing procedure for its secondary endpoints: hip fracture, then nonvertebral fracture. The trial reported a 40% reduction in hip fractures (0.7% vs 1.2%; p = 0.04) and a 20% reduction in nonvertebral fractures (6.5% vs 8.0%; p = 0.01). The hierarchical gate-keeping preserved the family-wise error rate at 0.05, meaning both secondary results carry valid p-values without alpha inflation.

Bone mineral density (BMD) changes were also reported: +9.2% at the lumbar spine and +6.0% at the total hip versus placebo over 36 months. These were not gated in the formal hierarchy and should be interpreted as supportive rather than confirmatory.

Why Placebo Instead of an Active Comparator?

By 2004 when FREEDOM began enrolling, alendronate had been on the market for nearly a decade. A reasonable question is why the trial used placebo rather than a head-to-head bisphosphonate comparison.

Three factors explain this. First, at the time of protocol design, no regulatory pathway required active-comparator fracture trials for osteoporosis; the FDA guidance for anti-resorptive agents accepted placebo-controlled fracture data. Second, a placebo comparison requires a smaller sample size to detect a treatment effect, reducing cost and timeline. Third, an active-comparator trial would have needed 15,000 to 20,000 patients to demonstrate non-inferiority or superiority against alendronate, given the relatively low absolute fracture rates in treated populations.

The trade-off is that FREEDOM cannot answer whether denosumab is better than bisphosphonates for fracture prevention. That question remained open until indirect comparisons and network meta-analyses emerged years later, and it still lacks a definitive head-to-head fracture RCT.

Statistical Design and Power

The trial was powered to detect a 50% reduction in new vertebral fractures assuming a 3-year placebo rate of approximately 7%. With 3,900 patients per arm, this provided over 90% power at a two-sided alpha of 0.05. The observed placebo rate of 7.2% aligned almost exactly with the planning assumption, and the observed effect size (68%) exceeded the minimum detectable difference.

The primary analysis used a logistic regression model with treatment group as the primary factor and baseline age, BMD T-score, and prevalent vertebral fracture status as covariates. A Cochran-Mantel-Haenszel test served as a sensitivity analysis. Both methods gave consistent results, reducing concern about model-dependence.

Missing data handling deserves scrutiny. Patients who discontinued treatment but remained in follow-up contributed to the intention-to-treat analysis. However, 24% of enrolled patients did not complete the 36-month visit. The published analysis used a modified ITT population, defined as all randomized patients who received at least one dose and had a baseline plus at least one post-baseline vertebral assessment. This excludes patients who dropped out before their first follow-up X-ray. Whether these early dropouts differed systematically from completers is not fully addressed.

Estimand Considerations

The ICH E9(R1) estimand framework was finalized a decade after FREEDOM, so the trial was not designed with explicit estimand language. Retrospectively, the primary analysis most closely approximates a treatment policy estimand: the effect of being assigned to denosumab regardless of adherence or discontinuation.

However, the modified ITT population introduces a wrinkle. By requiring at least one post-baseline X-ray, the analysis excludes very early dropouts. A pure treatment-policy estimand would include all randomized patients. The practical impact is likely small, since early dropouts represented a small fraction, but it means the headline result technically applies to patients who stayed long enough to be assessed, not to all patients who received a prescription.

For clinicians, the more relevant question is often the "while on treatment" estimand: what happens as long as the patient keeps taking the drug? FREEDOM's per-protocol analysis, restricted to patients who received all six injections, showed a slightly larger effect, consistent with the idea that adherent patients benefit more. The gap between ITT and per-protocol results was modest, suggesting that the ITT number is a reasonable approximation of the on-treatment benefit.

Safety and What the Design Could Not Catch

FREEDOM's 36-month duration was adequate for fracture efficacy but short for rare adverse events. Osteonecrosis of the jaw (ONJ) and atypical femoral fractures (AFF), two concerns that later became central to the denosumab safety conversation, were not observed in FREEDOM. ONJ events emerged in the open-label FREEDOM Extension that followed patients for up to 10 years, with a cumulative incidence of 5.2 per 10,000 patient-years.

The trial also did not assess rebound vertebral fractures after discontinuation, a risk that was only characterized years later when extension data and post-marketing studies showed that stopping denosumab leads to rapid bone loss and a cluster of vertebral fractures. The 2017 European Medicines Agency safety review and subsequent ASBMR position paper now recommend transitioning to a bisphosphonate after stopping denosumab, a clinical reality that FREEDOM's design could not anticipate.

Limitations the Authors Acknowledged

The original manuscript lists several limitations: the exclusion of women with very low T-scores (<−4.0), the lack of an active comparator, the 36-month duration being insufficient for rare events, and the requirement for calcium/vitamin D supplementation limiting generalizability to unsupplemented populations. These are genuine constraints, not boilerplate.

One limitation the authors did not emphasize: the trial enrolled exclusively postmenopausal women. Denosumab was later approved for men with osteoporosis and for glucocorticoid-induced osteoporosis based on separate, smaller trials, but FREEDOM itself provides no direct evidence for those populations.

What This Means for Clinical Practice

FREEDOM established denosumab as the first RANKL inhibitor with phase 3 fracture data, earning FDA approval in 2010 under the brand name Prolia. Its methodological rigor, central radiographic reading, hierarchical endpoint testing, and large sample size, makes the vertebral fracture result highly credible. The absence of an active comparator, the modified ITT population, and the 36-month ceiling on safety data are the main methodological gaps that subsequent studies have partially, but not completely, filled.

Frequently asked questions

›

References

Cummings SR, San Martin J, McClung MR, et al. Denosumab for prevention of fractures in postmenopausal women with osteoporosis. N Engl J Med. 2009;361(8):756-765. PubMed
Bone HG, Wagman RB, Brandi ML, et al. 10 years of denosumab treatment in postmenopausal women with osteoporosis: results from the phase 3 randomised FREEDOM trial and open-label extension. Lancet Diabetes Endocrinol. 2017;5(7):513-523. PubMed
Tsourdi E, Langdahl B, Cohen-Solal M, et al. Discontinuation of denosumab therapy for osteoporosis: a systematic review and position statement by ECTS. Bone. 2017;105:11-17. PubMed
FDA Approval Letter for Prolia (denosumab). June 1, 2010. FDA
Camacho PM, Petak SM, Binkley N, et al. American Association of Clinical Endocrinologists/American College of Endocrinology clinical practice guidelines for the diagnosis and treatment of postmenopausal osteoporosis, 2020 update. Endocr Pract. 2020;26(Suppl 1):1-46. PubMed