Honest Criticisms and Limitations of the FREEDOM Trial

At a glance

| Parameter | Detail | |-----------|--------| | N | 7,868 postmenopausal women | | Intervention | Denosumab 60 mg SC every 6 months | | Comparator | Placebo (both groups received calcium + vitamin D) | | Duration | 36 months | | Primary endpoint | New vertebral fractures at 36 months | | Key result | 68% relative risk reduction in new vertebral fractures (2.3% vs 7.2%) | | Sponsor | Amgen |

Enrollment Criteria Created a Narrow Patient Pool

The FREEDOM trial enrolled postmenopausal women aged 60 to 90 with lumbar spine or total hip T-scores between -2.5 and -4.0 (Cummings et al., NEJM 2009). That window sounds broad, but the exclusions tell the real story.

Women with severe osteoporosis (T-score below -4.0) were excluded. So were women currently on bisphosphonates, those with more than two prevalent vertebral fractures, and anyone with metabolic bone disease other than osteoporosis. Patients on systemic glucocorticoids, those with recent hip fracture, and women with significant renal impairment were also excluded.

This means the trial population was relatively "clean." They had osteoporosis but not the most severe forms. They were treatment-naive or had washed out prior therapy. They did not carry the comorbidity burden typical of 75-year-old women presenting in clinical practice with fragility fractures.

The median age was 72, and mean lumbar T-score was approximately -2.8. Fewer than 24% had a prevalent vertebral fracture at baseline. In real-world osteoporosis clinics, the population skews older, sicker, and more heavily pre-treated.

The 3-Year Window Was Enough for Approval, Not for Answers

Three years is a standard regulatory timeframe for fracture-reduction trials. It satisfied FDA requirements for osteoporosis drug approval. But osteoporosis treatment is lifelong, and the most concerning adverse events associated with anti-resorptive therapy (osteonecrosis of the jaw, atypical femoral fractures) emerge after 3 to 5 years of exposure.

In the original 3-year dataset, no cases of osteonecrosis of the jaw (ONJ) and no confirmed atypical femoral fractures (AFF) were reported in the denosumab arm (Cummings et al., NEJM 2009). This absence was frequently cited in early marketing. However, both complications surfaced in the FREEDOM Extension study at years 4 through 10, with ONJ incidence reaching approximately 5.2 per 10,000 patient-years by the seventh year of continuous use (Bone et al., Lancet Diabetes Endocrinol 2017).

The 3-year trial was not designed or powered to detect these rare but serious events. Citing their absence in FREEDOM as evidence of safety requires ignoring statistical power limitations the authors themselves acknowledged.

Statistical Design: What the Numbers Can and Cannot Say

The primary analysis used a pre-specified logistic regression model. The 68% relative risk reduction is the headline figure, but the absolute risk reduction was 4.8 percentage points (7.2% placebo vs 2.3% denosumab) for new vertebral fractures. The number needed to treat (NNT) was approximately 21 over 3 years.

For hip fracture, the secondary endpoint, the trial reported a 40% relative risk reduction. But the absolute numbers were small: 1.2% placebo vs 0.7% denosumab, yielding a confidence interval that barely excluded 1.0 (HR 0.60, 95% CI 0.37-0.97, p=0.04). With a p-value that close to the threshold and no adjustment for multiple comparisons on secondary endpoints, the hip fracture finding sits on fragile statistical ground.

The authors note that the trial was "not powered to detect differences in hip fracture" (Cummings et al., NEJM 2009). Post-hoc subgroup analyses of hip fracture in patients over 75 showed stronger effects, but these carry the standard caveats of exploratory analyses.

| Outcome | Denosumab | Placebo | RRR | ARR | NNT (3yr) | |---------|-----------|---------|-----|-----|-----------| | New vertebral fracture | 2.3% | 7.2% | 68% | 4.8% | ~21 | | Hip fracture | 0.7% | 1.2% | 40% | 0.5% | ~200 | | Non-vertebral fracture | 6.5% | 8.0% | 20% | 1.5% | ~67 |

The non-vertebral fracture reduction (20%) was modest and included fractures at sites where bone mineral density gains from denosumab are smaller.

Generalizability Gaps

Several populations were systematically excluded from FREEDOM yet commonly receive denosumab in practice:

Men. The trial enrolled only women. Male osteoporosis affects approximately 2 million men in the US. Denosumab's approval for men came from a separate, smaller trial (ADAMO, N=242).

Glucocorticoid-induced osteoporosis. Patients on systemic corticosteroids were excluded. This is one of the most common reasons for secondary osteoporosis and a frequent indication for denosumab prescribing.

Severe renal impairment. Patients with eGFR <30 were excluded. Denosumab is often chosen specifically because it does not require renal dose adjustment (unlike bisphosphonates), yet the safety data in advanced CKD came from post-marketing, not FREEDOM.

Sequential therapy patients. Women who had recently used bisphosphonates were excluded or required a washout. In practice, many patients switch from a bisphosphonate to denosumab due to intolerance or inadequate response. FREEDOM cannot speak to this transition population.

The American Association of Clinical Endocrinology 2020 guidelines acknowledge these gaps when positioning denosumab recommendations, noting that evidence for several common clinical scenarios derives from observational data rather than RCTs.

Industry Sponsorship and Conflict of Interest

Amgen funded FREEDOM, designed the protocol, collected the data, performed the statistical analysis, and employed several of the authors. The corresponding author (Steven Cummings) disclosed consulting fees from Amgen. Of the 12 authors, 8 reported financial ties to the sponsor.

This does not invalidate the results. Large fracture-reduction trials are expensive, and industry sponsorship is the norm. But the degree of sponsor involvement in data handling merits scrutiny. Independent academic statisticians did not perform the primary analysis. The data were not housed at an independent academic coordinating center.

A 2012 meta-analysis in the BMJ noted that industry-sponsored osteoporosis trials tend to report larger effect sizes than investigator-initiated studies, even when studying the same drug class (Defined by Jorgensen et al.). FREEDOM's effect sizes are consistent with what subsequent real-world data have shown, which lends credibility. Still, the structural conflicts deserve transparent disclosure in any evidence appraisal.

The Discontinuation Problem FREEDOM Did Not Address

FREEDOM was designed as a continuous-treatment trial. All participants received denosumab (or placebo) for the full 36 months. The protocol did not include a discontinuation arm or a treatment-holiday analysis.

This matters because denosumab's mechanism (RANKL inhibition) is fully reversible. When treatment stops, bone turnover markers rebound above baseline within 6 months, and BMD returns to pre-treatment levels within 12 to 24 months. Vertebral fracture risk after discontinuation was documented in FREEDOM Extension off-treatment data, with multiple vertebral fractures clustering in the first year after stopping (Cummings et al., JBMR 2018).

The original trial gave no signal about this because nobody stopped. Clinicians adopting denosumab based solely on FREEDOM had no framework for discontinuation planning until years later, when the rebound fracture phenomenon was characterized through extension and observational studies.

What Post-Publication Commentary Raised

Letters to the editor following the 2009 publication raised several points:

1. Infection signal. The denosumab group showed numerically higher rates of serious infections requiring hospitalization (4.1% vs 3.4%). The authors attributed this to chance. Subsequent meta-analyses have not confirmed a statistically significant infection risk, but the biological plausibility (RANKL is expressed on immune cells) keeps this concern active in pharmacovigilance.

2. Eczema and dermatologic events. Skin reactions were more common with denosumab (3.0% vs 1.7%). This was confirmed in post-marketing surveillance and is now listed in the Prolia prescribing information.

3. Cardiovascular events. Some correspondents noted a numerical imbalance in cardiovascular serious adverse events. Neither the original trial nor subsequent analyses have confirmed a causal relationship, but the trial was not powered or adjudicated for cardiovascular outcomes.

4. Pancreatitis. A small numerical excess of pancreatitis cases appeared. This remains in the label's adverse-events section but has not been substantiated as a class effect.

Comparison to Competing Evidence

When FREEDOM published, the bisphosphonate evidence base (FIT for alendronate, HORIZON for zoledronic acid) spanned longer durations and included extension data. The HORIZON trial demonstrated vertebral fracture reduction of 70% at 3 years with IV zoledronic acid, a nearly identical effect size, but with 6 years of extension data already emerging.

Critics argued that FREEDOM's headline result, while impressive, did not demonstrate superiority over existing bisphosphonates. No head-to-head fracture trial between denosumab and any bisphosphonate has been completed. The positioning of denosumab as "first-line" rests partly on its favorable dosing (twice-yearly injection vs weekly oral pill), partly on its use in renal impairment, and partly on FREEDOM's large N. But the fracture-reduction evidence is comparable, not superior, to the bisphosphonate class.

Summary of Key Limitations

| Limitation | Clinical Implication | |-----------|---------------------| | Treatment-naive, moderate-severity population | May overestimate benefit in sicker, pre-treated patients | | 3-year duration | Underpowered for rare long-term AEs (ONJ, AFF) | | No discontinuation arm | Missed the rebound fracture signal entirely | | Women only | Cannot directly inform male osteoporosis management | | Excluded GC-induced and CKD patients | Gaps in two of the most common prescribing scenarios | | Full sponsor control of data | No independent academic analysis | | Hip fracture barely significant | Marginal p-value, not powered for this endpoint |

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. 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
2. 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
3. Cummings SR, Ferrari S, Eastell R, et al. Vertebral fractures after discontinuation of denosumab: a post hoc analysis of the randomized placebo-controlled FREEDOM trial and its extension. J Bone Miner Res. 2018;33(2):190-198. PubMed
4. Prolia (denosumab) Prescribing Information. Amgen Inc. Revised 2020. FDA Label
5. 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