This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
DIRECT was a randomized, quadruple-masked, phase 3 parallel-group trial in patients with osteoporosis. The registered primary endpoint assessed the incidence of new or worsening vertebral fractures from baseline to 24 months, comparing denosumab with placebo.
| Feature | DIRECT |
|---|---|
| Phase | Phase 3 |
| Condition | Osteoporosis |
| Design | Randomized, parallel-group, quadruple-masked |
| Primary purpose | Treatment |
| Enrollment | 1262 |
| Arms | 3 |
| Interventions | Denosumab, placebo, alendronate sodium hydrate |
| Primary endpoint | Incidence of new or worsening vertebral fractures in osteoporotic subjects treated with denosumab compared to placebo |
| Primary endpoint time frame | Baseline to 24 months |
| Primary statistical method | Log-rank test |
| Primary effect measure | Hazard ratio |
| Hypothesis type | Superiority |
| Status | Completed |
| Lead sponsor | Daiichi Sankyo Co., Ltd. |
2. Clinical Question
The central statistical question was whether treatment with denosumab changed the time-to-event incidence of new or worsening vertebral fractures over the registered 24-month period compared with placebo in osteoporotic patients.
Population
Osteoporotic subjects enrolled in the DIRECT phase 3 trial.
Intervention
Denosumab.
Comparator
Placebo.
Primary question
Does denosumab reduce the hazard of a new or worsening vertebral fracture relative to placebo from baseline to 24 months?
The trial also included an alendronate sodium hydrate arm. The posted primary statistical analysis, however, reports the primary vertebral-fracture comparison between denosumab and placebo. The additional posted analyses of non-vertebral and hip fractures likewise compare denosumab with placebo.
3. Trial Design
Denosumab
- Denosumab was one of three registered interventions.
- The primary statistical comparison was denosumab versus placebo.
- Primary outcome: new or worsening vertebral fractures.
Placebo
- Placebo was one of three registered interventions.
- It served as the comparator for the posted primary and secondary statistical analyses.
- Primary outcome: new or worsening vertebral fractures.
Alendronate sodium hydrate
- Alendronate sodium hydrate was the third registered intervention.
- The posted statistical analyses summarized here do not provide a corresponding effect estimate for this arm.
Three-arm randomized comparison
- Total enrollment: 1262.
- Parallel-group design.
- Quadruple masking.
4. Endpoints
| Endpoint role | Registered outcome | Time frame | Statistical analysis posted |
|---|---|---|---|
| Primary | Incidence of New or Worsening Vertebral Fractures in Osteoporotic Subjects Treated With Denosumab Compared to Placebo | Baseline to 24 months | Log-rank test; HR reported |
| Secondary | The Percentage of Non-vertebral Fractures | Baseline to 24 Months | Log-rank test; P-value reported |
| Secondary | Percentage of Participants With Hip Fractures in Osteoporotic Participants Treated With Denosumab Compared to Treatment With Placebo. | Baseline to 24 Months | Log-rank test; P-value reported |
Although the secondary endpoint names use percentage terminology, the registry's posted analyses classify both secondary outcomes as time-to-event endpoints and analyze the denosumab-versus-placebo comparisons using the log-rank test.
5. Statistical Methodology
Time-to-event analysis
The primary endpoint was analyzed as a time-to-event outcome over baseline to 24 months. This is important statistically because the analysis is concerned not only with whether a vertebral fracture occurred, but with the timing of the event during follow-up.
Log-rank test
The posted primary analysis used the log-rank test to compare the time-to-event distributions for denosumab and placebo. The same method was used for the two posted secondary fracture analyses.
The log-rank test uses information from the sequence of events and the numbers of participants at risk over follow-up. It is therefore different from a simple comparison of percentages at a single time point.
Hazard ratio
The primary effect measure was the hazard ratio. The posted estimate compares the estimated instantaneous event rate for denosumab with that for placebo over the analyzed time-to-event follow-up.
An HR below 1 indicates a lower estimated instantaneous event rate in the denosumab group relative to placebo under the statistical model associated with the reported analysis.
Superiority hypothesis
The registered hypothesis type for the primary and posted secondary analyses was superiority. Thus, the statistical question was whether the denosumab and placebo time-to-event distributions differed in the direction specified by the treatment comparison, rather than whether denosumab met a non-inferiority margin.
Confidence interval
The primary analysis reported a two-sided 95% confidence interval for the hazard ratio. Confidence intervals provide information about the statistical precision of an estimated treatment effect and should be considered together with the point estimate rather than treating the point estimate as exact.
6. Primary Result: New or Worsening Vertebral Fractures
The primary endpoint was the incidence of new or worsening vertebral fractures from baseline to 24 months in osteoporotic subjects treated with denosumab compared with placebo. The registry reports a log-rank analysis with a hazard ratio of 0.343.
Hazard ratio for new or worsening vertebral fractures
95% CI: 0.194–0.606 · P = 0.0001
Denosumab vs placebo · Two-sided 95% CI · Superiority hypothesis
| Primary endpoint | Denosumab vs placebo |
|---|---|
| Outcome | Incidence of new or worsening vertebral fractures |
| Time frame | Baseline to 24 months |
| Endpoint type | Time-to-event |
| Statistical method | Log-rank test |
| Effect measure | Hazard ratio |
| Hazard ratio | 0.343 |
| 95% CI | 0.194–0.606 |
| P-value | 0.0001 |
| Hypothesis | Superiority |
The reported HR of 0.343 means that the estimated instantaneous rate of a new or worsening vertebral fracture in the denosumab group was 0.343 times the corresponding rate in the placebo group under the time-to-event analysis. Expressed as a relative hazard interpretation, this corresponds to an estimated 65.7% lower hazard because 1 − 0.343 = 0.657.
The HR does not mean that 65.7% of participants avoided fracture, nor does it mean that the probability of fracture was reduced by exactly 65.7% for every participant. A hazard ratio is a relative time-to-event measure, not an absolute risk difference or a treatment effect experienced identically by every individual.
The 95% CI of 0.194–0.606 describes uncertainty around the estimated hazard ratio under the statistical framework used for the analysis. It indicates that the reported point estimate should not be interpreted as a perfectly precise quantity. The interval is entirely below 1, which is consistent with the direction of the reported superiority result.
The P-value of 0.0001 addresses evidence against the null hypothesis represented by the statistical test. It does not measure the magnitude of the treatment effect. The magnitude is described by the HR and its confidence interval.
7. Secondary Result: Non-vertebral Fractures
The first posted secondary analysis evaluated the percentage of non-vertebral fractures from baseline to 24 months. The registry classifies the endpoint as time-to-event and reports a log-rank comparison between denosumab and placebo.
Log-rank comparison
Denosumab vs placebo · Baseline to 24 Months
| Secondary endpoint | Reported analysis |
|---|---|
| Outcome | The Percentage of Non-vertebral Fractures |
| Time frame | Baseline to 24 Months |
| Endpoint type | Time-to-event |
| Comparison | Denosumab vs placebo |
| Method | Log Rank |
| Effect measure | Not reported in the posted analysis |
| P-value | 0.9951 |
| Hypothesis type | Superiority |
The reported P-value of 0.9951 does not provide evidence of a statistically detectable difference between the denosumab and placebo time-to-event distributions under this log-rank analysis. Importantly, a P-value near 1 is not a measure of treatment equivalence and does not establish that the two groups have identical fracture risks.
The registry does not provide a hazard-ratio estimate or confidence interval for this secondary endpoint in the posted statistical analysis. Consequently, the P-value alone does not describe the magnitude or precision of a potential treatment difference.
This distinction is important: a nonsignificant hypothesis test and an equivalence conclusion are different statistical claims. Demonstrating equivalence or non-inferiority requires a prespecified margin and an analysis designed for that question; the registered hypothesis type here is superiority.
8. Secondary Result: Hip Fractures
The second posted secondary analysis evaluated the percentage of participants with hip fractures in osteoporotic participants treated with denosumab compared with placebo over baseline to 24 months. The registry classifies this endpoint as time-to-event and reports a log-rank comparison.
Log-rank comparison
Denosumab vs placebo · Baseline to 24 Months
| Secondary endpoint | Reported analysis |
|---|---|
| Outcome | Percentage of Participants With Hip Fractures in Osteoporotic Participants Treated With Denosumab Compared to Treatment With Placebo. |
| Time frame | Baseline to 24 Months |
| Endpoint type | Time-to-event |
| Comparison | Denosumab vs placebo |
| Method | Log Rank |
| Effect measure | Not reported in the posted analysis |
| P-value | 0.1568 |
| Hypothesis type | Superiority |
The P-value of 0.1568 does not meet a conventional 0.05 threshold for statistical significance. More importantly, the P-value alone does not establish that denosumab and placebo have the same hip-fracture hazard.
Because the registry's posted analysis does not provide a hazard ratio or confidence interval for this endpoint, the statistical evidence cannot be summarized from the posted result as a quantified relative treatment effect. The result is therefore best understood as a log-rank hypothesis-test result rather than as a measure of effect magnitude.
The endpoint is also clinically distinct from the primary vertebral-fracture endpoint. A treatment effect for vertebral fractures need not imply an identical effect for every fracture subtype, and each endpoint should be interpreted according to its own observed analysis.
9. Safety Results
The registry reports serious adverse events by treatment arm. These counts are presented as affected participants over participants at risk.
| Treatment arm | Serious adverse events | Affected / at risk |
|---|---|---|
| Denosumab | Serious adverse events | 65 / 475 |
| Placebo | Serious adverse events | 68 / 481 |
| Alendronate | Serious adverse events | 30 / 242 |
The reported serious-adverse-event counts should be interpreted separately from the efficacy analyses. The efficacy endpoint was a time-to-event fracture outcome analyzed by a log-rank test, whereas the safety information reported here is a count of participants affected among those at risk. The two types of information answer different questions.
10. Statistical Methods Explained
Why was a log-rank test used?
The primary endpoint was a time-to-event outcome assessed over baseline to 24 months. A log-rank test is designed to compare the survival or event-free distributions of two groups while incorporating the timing of events during follow-up. That makes it more informative for this type of endpoint than simply comparing whether an event occurred by a single fixed date.
What does an HR of 0.343 mean?
An HR of 0.343 indicates that the estimated instantaneous vertebral-fracture rate in the denosumab group was 0.343 times that of the placebo group under the reported time-to-event analysis. Equivalently, the complementary relative-hazard interpretation is approximately a 65.7% lower estimated hazard. This is not the same as saying that 65.7% of participants avoided fracture or that every patient's individual probability was reduced by 65.7%.
Why does the confidence interval matter?
The primary 95% confidence interval, 0.194–0.606, conveys the uncertainty surrounding the HR estimate. A point estimate such as 0.343 is only one estimate from the observed trial. The interval communicates how precisely the treatment effect was estimated within the statistical framework used for the analysis.
Why does the P-value not measure treatment effect size?
The primary P-value of 0.0001 describes the strength of evidence against the null hypothesis under the log-rank test. It does not tell the reader that the treatment effect is "large" or "small." Effect magnitude is better described by the hazard ratio, while its uncertainty is described by the confidence interval.
Why can a nonsignificant P-value not prove equivalence?
The secondary non-vertebral-fracture analysis has a P-value of 0.9951, and the hip-fracture analysis has a P-value of 0.1568. Neither result establishes equivalence. A superiority test asks whether the groups differ; failure to reject its null hypothesis is not the same statistical question as demonstrating that two treatments are sufficiently similar within a prespecified equivalence margin.
Why is censoring important in time-to-event analysis?
Not every participant necessarily experiences the fracture event during the observation period. Time-to-event methods can incorporate participants whose event status is not observed through the entire period by using their available follow-up information. This is one reason the timing of follow-up and the handling of censoring are fundamental to interpretation of a log-rank analysis and hazard ratio.
11. Understanding the Primary Hazard Ratio
The primary HR of 0.343 indicates a substantially lower estimated instantaneous hazard of new or worsening vertebral fracture for denosumab relative to placebo under the reported analysis. The complementary calculation, 1 − 0.343, gives 0.657, or 65.7%, as a relative-hazard reduction interpretation.
The HR does not provide the absolute probability that a particular patient will experience a vertebral fracture. It also does not specify the absolute difference in event-free survival at 24 months, because those quantities require corresponding absolute event or survival estimates.
The 95% CI of 0.194–0.606 indicates that the estimated treatment effect has uncertainty around the point estimate. The entire interval lies below 1, which is consistent with the direction of the reported superiority result.
The P-value of 0.0001 is evidence against the null hypothesis used by the log-rank test. It should not be used as a substitute for the effect estimate or confidence interval.
12. Interpreting the Secondary P-values
| Endpoint | P-value | What the posted result tells us |
|---|---|---|
| New or worsening vertebral fractures | 0.0001 | The reported primary log-rank comparison provides strong evidence against the null hypothesis. |
| Non-vertebral fractures | 0.9951 | The reported log-rank comparison does not provide evidence of a statistically detectable difference under the superiority test. |
| Hip fractures | 0.1568 | The reported log-rank comparison does not meet a conventional 0.05 threshold for statistical significance. |
The three P-values should not be interpreted as if they were three measurements of the same endpoint. They correspond to distinct fracture outcomes. The primary vertebral-fracture result includes a hazard ratio and confidence interval, whereas the posted secondary analyses summarized here report P-values without corresponding effect estimates.
13. Multiplicity and Multiple Endpoints
DIRECT has one registered primary endpoint and two posted secondary analyses in the statistical information summarized here. The distinction between primary and secondary outcomes matters because a trial can evaluate several outcomes while assigning different inferential roles to them.
| Endpoint | Role | Posted statistical information |
|---|---|---|
| New or worsening vertebral fractures | Primary | Log-rank test; HR 0.343; 95% CI 0.194–0.606; P = 0.0001 |
| Non-vertebral fractures | Secondary | Log-rank test; P = 0.9951 |
| Hip fractures | Secondary | Log-rank test; P = 0.1568 |
When multiple outcomes are examined, interpretation should preserve the prespecified hierarchy of the trial rather than treating every statistical test as though it carried the same confirmatory status. The registry identifies the vertebral-fracture analysis as the primary endpoint.
14. Randomization and Internal Validity
Randomization is central to the interpretation of a treatment comparison because it is intended to balance measured and unmeasured prognostic factors between treatment groups in expectation. DIRECT is registered as randomized and parallel-group, with quadruple masking.
Randomization
Random allocation creates the basis for comparing outcomes between treatment groups without relying solely on adjustment for measured baseline variables.
Masking
Quadruple masking reduces opportunities for knowledge of treatment assignment to influence trial conduct or assessment.
Parallel design
Participants remain in their randomized treatment groups rather than serving as their own sequential treatment controls.
Superiority framework
The registered hypothesis asks whether denosumab and placebo differ, rather than whether their effects fall within a predefined non-inferiority or equivalence margin.
15. What Can and Cannot Be Learned From the Posted Analyses
The primary analysis supports a quantified time-to-event treatment comparison because the registry reports an HR, a 95% confidence interval, and a P-value. The two secondary analyses provide hypothesis-test results but do not include corresponding HR estimates or confidence intervals in the posted statistical analyses summarized here.
| Question | What the posted analysis provides |
|---|---|
| Is there a reported primary treatment-effect estimate? | Yes. HR = 0.343. |
| Is uncertainty around the primary HR reported? | Yes. Two-sided 95% CI = 0.194–0.606. |
| Is a primary hypothesis-test result reported? | Yes. P = 0.0001. |
| Is an HR reported for non-vertebral fractures? | No posted HR is included in the summarized analysis. |
| Is an HR reported for hip fractures? | No posted HR is included in the summarized analysis. |
| Are P-values reported for the secondary analyses? | Yes. 0.9951 for non-vertebral fractures and 0.1568 for hip fractures. |
This distinction is statistically important. A P-value can indicate the result of a hypothesis test, but it does not by itself quantify how large a treatment effect is. For the primary endpoint, the HR and confidence interval supply that additional information.
16. Important Limitations and Interpretation Issues
- Limited effect information for secondary endpoints: the posted secondary analyses provide P-values but no corresponding hazard ratios or confidence intervals in the statistical analyses summarized here.
- Time-to-event interpretation: the primary and secondary outcomes are classified as time-to-event endpoints, so their analysis depends on the timing of events and available follow-up rather than only on event counts at one fixed time.
- Hazard ratio interpretation: the primary HR is a relative time-to-event measure. It should not be translated directly into an absolute fracture probability or an individual-level treatment effect.
- Confidence interval: the primary confidence interval communicates uncertainty around the HR but does not describe the range of effects that individual patients will experience.
- Multiple endpoints: the trial contains a primary endpoint and secondary fracture outcomes. Their inferential roles should remain distinct.
- Missing absolute event estimates: the posted statistical analyses summarized here do not provide the absolute fracture-free probabilities needed to express the primary result as an absolute risk difference.
- Generalizability: the trial record describes a specific population of osteoporotic subjects and therefore the applicability of the findings depends on how closely other populations match the enrolled population.
17. Why This Trial Matters Statistically
DIRECT is a useful teaching example because it illustrates how a randomized clinical trial can evaluate fracture outcomes as time-to-event endpoints rather than reducing the analysis to a simple comparison of event percentages.
| Concept | How it appears in DIRECT |
|---|---|
| Randomization | Participants were randomized in a three-arm parallel-group phase 3 trial. |
| Masking | The registry describes the study as quadruple-masked. |
| Time-to-event endpoint | The primary vertebral-fracture outcome is classified as time-to-event. |
| Kaplan-Meier framework | Time-to-event fracture analyses are naturally interpreted through survival/event-free methods. |
| Log-rank test | The posted primary and secondary comparisons use the log-rank method. |
| Hazard ratio | The primary endpoint is summarized by an HR of 0.343. |
| Confidence interval | The primary HR has a two-sided 95% CI of 0.194–0.606. |
| P-value | The primary analysis reports P = 0.0001, with secondary P-values of 0.9951 and 0.1568. |
| Superiority testing | The primary and secondary analyses are registered as superiority hypotheses. |
| Safety analysis | Serious adverse events are reported separately by treatment arm. |
18. Clinical Biostats Statistical Perspective
Primary efficacy signal
The primary analysis reports an HR of 0.343 with a two-sided 95% CI of 0.194–0.606 and P = 0.0001 for new or worsening vertebral fractures.
Secondary outcomes
The posted secondary log-rank P-values are 0.9951 for non-vertebral fractures and 0.1568 for hip fractures.
Relative versus absolute effects
The primary HR quantifies a relative time-to-event comparison. It does not by itself provide an absolute fracture risk or absolute risk reduction.
Efficacy versus safety
The serious-adverse-event counts are a separate component of the evidence and should not be merged mathematically with the primary efficacy result.
19. Sources and Trial Record
- ClinicalTrials.gov: NCT00680953 — DIRECT.
- PubMed: PMID 24646104.
- PubMed: PMID 25403903.
- PubMed: PMID 32839843.
20. Related Tutorials
Learn more about the methods used in this trial:
21. Related Statistical Calculators
Continue through the Clinical Biostats knowledge graph
Connect the trial's endpoints and methods to deeper statistical tutorials and analysis tools.
22. Record Summary
DIRECT provides a clear example of a randomized phase 3 trial in which the primary clinical outcome was analyzed as a time-to-event endpoint. The registry reports a log-rank comparison of denosumab versus placebo for new or worsening vertebral fractures from baseline to 24 months, with an HR of 0.343, a two-sided 95% CI of 0.194–0.606, and a P-value of 0.0001. The two posted secondary fracture analyses use the same log-rank framework but report only P-values: 0.9951 for non-vertebral fractures and 0.1568 for hip fractures.
Statistically, the most important lesson is the distinction between a hazard ratio, a confidence interval, and a P-value. The HR describes the relative time-to-event effect, the confidence interval describes uncertainty around that estimate, and the P-value evaluates evidence against the null hypothesis. None of these measures alone supplies the complete clinical picture. Safety is also evaluated separately, with serious adverse events reported as 65/475 for denosumab, 68/481 for placebo, and 30/242 for alendronate.