What was the primary endpoint in HORIZON-PFT?

New morphometric vertebral fracture, defined as ≥20% height loss in any vertebral body on lateral spine radiographs using the Genant semiquantitative method. This captures both symptomatic and asymptomatic fractures.

How was blinding maintained with an IV infusion?

Both groups received IV infusions (zoledronic acid or normal saline). Radiograph readers at a central facility were blinded to allocation. However, acute-phase reactions in ~32% of active-treatment patients may have partially unblinded clinical-site staff and patients.

Why did the trial require vitamin D repletion before randomization?

Vitamin D deficiency can independently cause bone loss and would confound fracture outcomes. Mandatory repletion ensured the trial isolated zoledronic acid's antiresorptive effect from simple nutrient correction.

What does the 70% reduction actually mean in absolute terms?

The morphometric vertebral fracture rate was 3.3% in the zoledronic acid group vs 10.9% in the placebo group over 3 years. The absolute risk reduction was 7.6 percentage points, yielding a number needed to treat of approximately 13.

Was HORIZON-PFT an intention-to-treat analysis?

It used modified intention-to-treat, including all randomized patients who received at least one infusion and had evaluable baseline plus follow-up radiographs. About 7% of randomized patients were excluded from the primary analysis.

Did the trial compare zoledronic acid to oral bisphosphonates?

No. The comparator was IV placebo (normal saline). No head-to-head fracture trial comparing zoledronic acid to oral alendronate or risedronate has been completed.

How diverse was the study population?

Minimally. Approximately 98% of participants were white postmenopausal women. The trial excluded men, premenopausal women, and those with severe renal impairment (CrCl <30 mL/min).

What happened when patients stopped zoledronic acid after 3 years?

Extension data showed modest increases in bone turnover markers and a small rise in vertebral fracture risk, but no significant increase in hip fractures. This informed current drug-holiday protocols for patients at moderate risk.

How does the statistical gatekeeping procedure affect interpretation?

Secondary endpoints (hip fracture, nonvertebral fracture) were only formally testable because the primary endpoint reached significance first. All passed, preserving strong type I error control across the endpoint hierarchy.

Is the 70% reduction clinically meaningful for all patients?

The absolute benefit depends on baseline fracture risk. For women with prevalent vertebral fractures (higher baseline risk), the absolute reduction is larger. For women qualifying solely on T-score without prior fracture, the absolute benefit is smaller despite the same relative reduction.

Inside the HORIZON-PFT Methodology: What Most Summaries Skip

Q: What happened when patients stopped zoledronic acid after 3 years?

Extension data showed modest increases in bone turnover markers and a small rise in vertebral fracture risk, but no significant increase in hip fractures. This informed current drug-holiday protocols for patients at moderate risk.

Q: How does the statistical gatekeeping procedure affect interpretation?

Secondary endpoints (hip fracture, nonvertebral fracture) were only formally testable because the primary endpoint reached significance first. All passed, preserving strong type I error control across the endpoint hierarchy.

Q: Is the 70% reduction clinically meaningful for all patients?

The absolute benefit depends on baseline fracture risk. For women with prevalent vertebral fractures (higher baseline risk), the absolute reduction is larger. For women qualifying solely on T-score without prior fracture, the absolute benefit is smaller despite the same relative reduction.

By HealthRX Medical Team

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

At a glance

| Parameter | Detail | |-----------|--------| | N | 7,765 postmenopausal women | | Intervention | Zoledronic acid 5 mg IV once yearly (3 infusions total) | | Comparator | Matching IV placebo (normal saline) | | Duration | 3 years | | Primary endpoint | New morphometric vertebral fracture | | Key result | 70% relative risk reduction (3.3% vs 10.9%; absolute risk reduction 7.6%) | | Registration | NCT00049829 |

The Randomization Strategy Most Summaries Ignore

HORIZON-PFT enrolled postmenopausal women aged 65 to 89 with either a femoral neck T-score of <-2.5 (with or without existing vertebral fractures) or a T-score of <-1.5 with radiographic evidence of at least two mild or one moderate existing vertebral fracture. This dual-pathway entry criterion matters because it created a heterogeneous population spanning both primary and secondary prevention of fractures.

Randomization was 1:1, stratified by three variables: number of prevalent vertebral fractures (0, 1, 2, or 3+), years since menopause (<5 vs ≥5 years), and region (North America, Europe, Oceania, Latin America). The stratification by prevalent fracture count is particularly important. Women with existing fractures have roughly four times the risk of new fractures compared to those without, so balancing this variable protects against confounding that could distort the treatment effect in either direction (Black et al., 2007).

Background Therapy: The Confounder That Wasn't Hidden

All participants received calcium (1,000 to 1 to 500 mg daily) and vitamin D (400 to 1 to 200 IU daily). This was not a minor detail. Vitamin D deficiency accelerates bone loss and can independently increase fracture risk. The protocol required a loading dose of vitamin D (50,000 to 125 to 000 IU orally or intramuscularly) before randomization if 25-hydroxyvitamin D levels were below 15 ng/mL.

This mandatory repletion strategy means the trial tested zoledronic acid on top of adequate calcium and vitamin D, not in isolation. Clinicians extrapolating the 70% reduction to patients with untreated vitamin D deficiency should note that the trial's internal validity depends on this co-intervention (Black et al., 2007).

Participants already on osteoporosis therapies were excluded, but the protocol permitted washout: prior bisphosphonate use was allowed if stopped ≥12 months before enrollment for oral agents and ≥2 years for IV agents. Prior raloxifene or calcitonin was allowed with a 6-month washout.

The Blinding Problem With IV Infusions

Double-blinding an annual IV infusion presents a unique challenge. Zoledronic acid causes an acute-phase reaction (fever, myalgia, arthralgia) in approximately 30-40% of patients after the first infusion. This is a well-documented class effect of nitrogen-containing bisphosphonates given intravenously.

The HORIZON investigators addressed potential unblinding by having the outcome assessors (radiologists reading vertebral films) remain separate from the clinical sites where infusions occurred. Patients and site investigators, however, could potentially infer allocation based on post-infusion symptoms. The 2007 publication reports that formal assessment of blinding integrity was not performed, a limitation worth noting when interpreting subjective secondary endpoints like back pain.

For the primary endpoint (morphometric vertebral fracture), this potential unblinding is less concerning because fracture identification relied on standardized quantitative morphometry performed by a centralized reading facility blinded to treatment assignment.

The Primary Endpoint: Morphometric vs Clinical Fracture

This is where most summaries fail readers. The primary endpoint was new morphometric vertebral fracture, defined using the Genant semiquantitative method as a ≥20% reduction in anterior, middle, or posterior vertebral body height on lateral thoracic and lumbar spine radiographs compared with baseline.

This definition captures fractures that may be clinically silent. Approximately two-thirds of vertebral fractures identified radiographically never come to clinical attention. The distinction matters for clinical interpretation:

| Endpoint | Zoledronic acid | Placebo | Relative reduction | |----------|----------------|---------|-------------------| | Morphometric vertebral fracture | 3.3% | 10.9% | 70% (95% CI: 62-76%) | | Clinical vertebral fracture | 0.5% | 2.6% | 77% | | Hip fracture | 1.4% | 2.5% | 41% | | Any nonvertebral fracture | 8.0% | 10.7% | 25% |

The hierarchy between morphometric and clinical fractures is not trivial. The 77% reduction in clinical vertebral fractures (those causing symptoms prompting medical attention) was a secondary endpoint, not the primary one. The FDA's labeling for Reclast references morphometric fractures because that was the pre-specified primary analysis (FDA Reclast Label).

Statistical Architecture: Gatekeeping and the mITT Population

The statistical analysis plan used a hierarchical (gatekeeping) testing procedure to control the family-wise type I error rate. The primary endpoint was tested first at α = 0.05. Only if significant could secondary endpoints (hip fracture, nonvertebral fracture, clinical vertebral fracture) be tested sequentially, each at α = 0.05.

The primary analysis population was modified intention-to-treat (mITT): all randomized patients who received at least one dose of study drug and had a baseline and at least one post-baseline vertebral radiograph. This excluded approximately 7% of randomized patients who lacked evaluable follow-up radiographs (Black et al., 2007).

The choice of mITT over strict ITT deserves scrutiny. Patients who dropped out before obtaining a follow-up radiograph were disproportionately from the zoledronic acid group (acute-phase reactions leading to early withdrawal). Excluding these patients could theoretically bias results in favor of treatment if those who left were sicker or more fracture-prone. The investigators addressed this with sensitivity analyses imputing missing data, which produced consistent results, but the gap between mITT and strict ITT remains a legitimate methodological critique.

The Estimand Framework Applied Retrospectively

HORIZON-PFT predates the ICH E9(R1) addendum on estimands (2019), but applying that framework retrospectively illuminates what question the trial actually answered:

Population: Postmenopausal women 65-89 with osteoporosis (T-score or fracture-defined)
Treatment: Zoledronic acid 5 mg IV yearly, assuming calcium/vitamin D repletion
Intercurrent events: Treatment discontinuation handled by mITT (de facto treatment-policy strategy for the primary endpoint)
Population-level summary: Risk difference at 3 years

The treatment-policy estimand means the primary result reflects the effect of assigning zoledronic acid, not the effect of completing all three infusions. With 83% of zoledronic acid patients and 85% of placebo patients completing the trial, adherence was high, so the per-protocol and ITT estimates converge closely here, unlike oral bisphosphonate trials where adherence typically drops below 50% by year two.

Comparator Choice and Its Clinical Implications

The placebo comparator (normal saline IV) was appropriate for regulatory approval but leaves a clinical question unanswered: how does annual IV zoledronic acid compare to weekly oral alendronate? No head-to-head fracture endpoint trial has been completed. The HORIZON design answers "does zoledronic acid reduce fractures compared to no antiresorptive?" but not "should clinicians prefer it over oral options?"

Indirect comparisons using the FIT trial (alendronate) suggest similar vertebral fracture reduction magnitudes (47% for alendronate at 3 years vs 70% for zoledronic acid), but cross-trial comparisons are unreliable due to different baseline risk populations and fracture definitions. The American Association of Clinical Endocrinology 2020 guidelines position zoledronic acid as preferred for high-risk patients partly based on adherence advantages rather than proven fracture superiority.

Limitations the Authors Acknowledged

The original publication and subsequent analyses identified several constraints:

Population homogeneity: 98% of participants were white. Fracture epidemiology and bisphosphonate metabolism may differ across racial groups.
Renal exclusion: Patients with creatinine clearance <30 mL/min were excluded. Post-marketing surveillance has documented rare nephrotoxicity, and the trial cannot inform safety in advanced CKD.
Duration ceiling: Three years of data. The optimal duration of therapy and the long-term safety profile (particularly regarding atypical femoral fractures and osteonecrosis of the jaw) required subsequent extension studies.
Acute-phase reaction: 31.6% of zoledronic acid patients experienced post-infusion symptoms after dose one, potentially compromising blinding and causing early dropout.
Concomitant vitamin D: The mandatory repletion protocol means the trial population may not reflect real-world patients who begin zoledronic acid without prior optimization.

What the Follow-Up Data Show

The HORIZON extension study followed a subset of patients to 6 years (3 years on zoledronic acid followed by 3 years of either continued treatment or placebo). Results published in the Journal of Bone and Mineral Research (2012) showed that discontinuation after 3 years led to a modest increase in vertebral fracture risk but not hip fracture risk, informing current "drug holiday" recommendations. The 2022 Endocrine Society guidelines cite this extension data when recommending reassessment after 3 to 6 years of IV bisphosphonate therapy.

Frequently asked questions

›

References

Black DM, Delmas PD, Eastell R, et al. Once-yearly zoledronic acid for treatment of postmenopausal osteoporosis. N Engl J Med. 2007;356(18):1809-1822. PubMed
FDA. Reclast (zoledronic acid) prescribing information. Revised 2018. FDA Label
Black DM, Reid IR, Boonen S, et al. The effect of 3 versus 6 years of zoledronic acid treatment of osteoporosis: a randomized extension to the HORIZON-Key Fracture Trial. J Bone Miner Res. 2012;27(2):243-254. PubMed
Black DM, Cummings SR, Karpf DB, et al. Randomised trial of effect of alendronate on risk of fracture in women with existing vertebral fractures (FIT). Lancet. 1996;348(9041):1535-1541. PubMed
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