This page separates reported trial results from statistical interpretation. The numerical results presented here are limited to the ClinicalTrials.gov record for D-CARE. The registry provides the official trial record.
1. Trial at a Glance
D-CARE was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating denosumab versus placebo as adjuvant treatment for women with high-risk early breast cancer receiving neoadjuvant or adjuvant therapy.
| Feature | D-CARE |
|---|---|
| Trial name | D-CARE |
| NCT identifier | NCT01077154 |
| Phase | Phase 3 |
| Status | Terminated |
| Therapeutic area | Oncology |
| Condition | Breast Cancer |
| Population description | Women with high-risk early breast cancer receiving neoadjuvant or adjuvant therapy |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 4,509 |
| Lead sponsor | Amgen |
| Sponsor type | Industry |
2. Clinical Question
The central statistical question was whether denosumab changed bone metastasis-free survival compared with placebo in the randomized population of women with high-risk early breast cancer receiving neoadjuvant or adjuvant therapy.
Population
Women with high-risk early breast cancer receiving neoadjuvant or adjuvant therapy.
Intervention
Denosumab.
Comparator
Placebo.
Primary question
Does denosumab improve bone metastasis-free survival relative to placebo?
3. Trial Design
Denosumab
- Denosumab
- Drug intervention
Placebo
- Placebo
- Drug/placebo intervention
4. Endpoints
The registry lists one primary endpoint, and it is a time-to-event endpoint. Five outcome measures and five statistical analyses are posted in the ClinicalTrials.gov record.
| Endpoint | Registry definition / time frame | Type |
|---|---|---|
| Bone Metastasis-free Survival (BMFS) | BMFS time was defined as the time interval from the randomization date to the first occurrence of bone metastasis or death from any cause, whichever came first. Participants last known to be alive with no bone metastasis were censored at their last assessment date, or at the primary analysis data cut-off date, whichever was first. Time frame: from randomization until the primary analysis data cut-off date of 31 August 2017. | Time-to-event |
| Disease-free Survival (DFS) | From randomization until the primary analysis data cut-off date of 31 August 2017. | Time-to-event |
| Disease-free Survival (DFS) in the Postmenopausal Subset | From randomization until the primary analysis data cut-off date of 31 August 2017. The analysis population was randomized participants postmenopausal at enrollment, as defined in the registry. | Time-to-event |
| Overall Survival | From randomization until the end of study; median (minimum, maximum) time on study was 72.7 (0, 92) and 72.3 (0, 92) months. | Time-to-event |
| Distant Recurrence-free Survival | From randomization until the primary analysis data cut-off date of 31 August 2017. | Time-to-event |
5. Statistical Methodology
Primary analysis population
The primary BMFS analysis was conducted in all randomized participants. This is an important feature of the analysis because treatment assignment, rather than treatment received, defines the primary efficacy population reported by the registry.
Stratified log-rank test
The registry reports a stratified log-rank test as the method for BMFS and the four posted secondary time-to-event analyses. The analysis notes state that the comparison was stratified by breast cancer therapy/LN status, hormone receptor status, HER-2 status, age, and geographic region.
The stratified log-rank test compares the observed timing of events between treatment groups while accounting for the prespecified stratification factors. It is designed for time-to-event data, where both event occurrence and follow-up time contribute to the comparison.
Stratified Cox proportional-hazards model
The registry states that the hazard ratio was based on a Cox proportional-hazards model stratified by the randomization stratification factors. The reported analysis notes state that a hazard ratio below 1 favors denosumab.
The hazard ratio is a relative, model-based measure of the event rate over time. It is not an absolute risk difference, a probability of benefit, or a statement that every participant experiences the same proportional change.
Confidence intervals
Each posted analysis reports a two-sided 95% confidence interval for the hazard ratio. The interval describes statistical uncertainty around the estimated treatment effect under the analysis model and sampling framework.
Superiority hypothesis
The hypothesis type reported for all five statistical analyses is superiority. Thus, the inferential question is whether the randomized groups differ in the time-to-event outcome, rather than whether one treatment is merely no worse than the other within a prespecified non-inferiority margin.
6. Results: Bone Metastasis-free Survival
BMFS is the single registered primary endpoint. The analysis used all randomized participants and compared placebo with denosumab using a stratified log-rank test. The effect measure was a hazard ratio from a stratified Cox proportional-hazards model.
Primary BMFS hazard ratio
95% CI: 0.82–1.14 · P = 0.70
Analysis population: all randomized participants · Two-sided 95% CI
| Primary endpoint | Analysis population | Method | Hazard ratio | 95% CI | P-value |
|---|---|---|---|---|---|
| Bone Metastasis-free Survival (BMFS) | All randomized participants | Stratified log-rank; stratified Cox model | 0.97 | 0.82–1.14 | 0.70 |
The estimated hazard ratio of 0.97 compares the instantaneous rate of the BMFS event between denosumab and placebo under the reported stratified Cox model. Because a hazard ratio below 1 favors denosumab in the registry's analysis notes, the point estimate is slightly below the null value of 1.
It does not mean that denosumab reduced the probability of bone metastasis or death by 3%, nor does it mean that individual participants experienced a 3% reduction in risk. A hazard ratio is not the same as a relative risk or an absolute risk difference.
The 95% CI of 0.82–1.14 shows that the estimate has appreciable statistical uncertainty. The interval includes 1, the value corresponding to equal estimated hazards under this measure.
The P-value of 0.70 is a measure of compatibility between the observed result and the null hypothesis under the specified testing framework. It is not a measure of effect size, clinical importance, or the probability that the null hypothesis is true.
Because this is a Cox-model hazard ratio, interpretation also depends on the model's time-to-event framework and the proportional-hazards assumption. The ClinicalTrials.gov record does not provide a separate assessment of that assumption.
7. Secondary Endpoint Results
Four secondary time-to-event analyses are posted in the registry data. Each uses all randomized participants except the specified postmenopausal subset analysis, and each compares placebo with denosumab using a stratified log-rank test and a hazard ratio from a stratified Cox model.
Disease-free Survival
DFS hazard ratio
95% CI: 0.91–1.19 · P = 0.57
The DFS hazard ratio of 1.04 is close to 1. Under the registry's convention that a hazard ratio below 1 favors denosumab, this point estimate is slightly above the null value.
The 95% CI of 0.91–1.19 includes 1, so the interval encompasses both a modestly lower and a modestly higher estimated hazard relative to placebo. The P-value of 0.57 should not be interpreted as the probability that there is no treatment effect; it quantifies evidence against the specified null hypothesis under the testing framework.
Disease-free Survival in the Postmenopausal Subset
Postmenopausal DFS hazard ratio
95% CI: 0.92–1.36 · P = 0.26
The estimated hazard ratio of 1.12 comes from the randomized participants who were postmenopausal at enrollment, using the registry's specified definition. The point estimate is above 1, whereas the registry defines values below 1 as favoring denosumab.
The 95% CI of 0.92–1.36 includes 1. Its width illustrates the uncertainty associated with estimating an effect in a subset rather than the complete randomized population. The P-value of 0.26 is not an effect-size measure and should not be used to compare the magnitude of this estimate with the overall DFS result by itself.
Overall Survival
Overall survival hazard ratio
95% CI: 0.83–1.22 · P = 0.94
Time frame: from randomization until the end of study
The overall-survival hazard ratio of 1.01 is very close to the null value of 1. The 95% CI of 0.83–1.22 spans both sides of 1, indicating uncertainty in the direction and magnitude of the estimated relative hazard.
The P-value of 0.94 indicates little evidence against the null hypothesis in this analysis. It does not establish that the two treatments are identical, and it does not measure whether any particular difference would be clinically important.
Distant Recurrence-free Survival
Distant recurrence-free survival hazard ratio
95% CI: 0.92–1.21 · P = 0.41
The distant recurrence-free survival hazard ratio of 1.06 is close to 1 and lies above the registry's favorable direction for denosumab. The 95% CI of 0.92–1.21 includes 1, while the P-value of 0.41 does not provide strong evidence against the null hypothesis under the reported analysis.
As with the other time-to-event results, the hazard ratio should not be translated directly into an absolute probability or an individual patient's risk.
8. Summary of Reported Time-to-Event Results
| Endpoint | Role | Population | HR | 95% CI | P-value |
|---|---|---|---|---|---|
| Bone Metastasis-free Survival | Primary | All randomized participants | 0.97 | 0.82–1.14 | 0.70 |
| Disease-free Survival | Secondary | All randomized participants | 1.04 | 0.91–1.19 | 0.57 |
| Disease-free Survival in the Postmenopausal Subset | Secondary | Randomized participants postmenopausal at enrollment | 1.12 | 0.92–1.36 | 0.26 |
| Overall Survival | Secondary | All randomized participants | 1.01 | 0.83–1.22 | 0.94 |
| Distant Recurrence-free Survival | Secondary | All randomized participants | 1.06 | 0.92–1.21 | 0.41 |
Across the five posted statistical analyses, the reported hazard-ratio estimates are all close to the null value of 1, and every reported two-sided P-value is greater than 0.05. The primary endpoint result is therefore best understood through its complete estimate, confidence interval, and testing result rather than through the P-value alone.
9. Serious Adverse Events
The ClinicalTrials.gov record reports serious adverse events by randomized arm using affected participants over the number at risk.
| Arm | Affected / at risk |
|---|---|
| Placebo | 675 / 2,218 |
| Denosumab | 702 / 2,241 |
These figures describe the number of affected participants and the corresponding numbers at risk as reported in the registry data. They are not themselves a formal hypothesis test or an adjusted comparison of safety risk.
10. Statistical Methods Explained
Why was a stratified log-rank test used?
A stratified log-rank test is appropriate for comparing randomized groups when the endpoint is time-to-event and participants may be censored before experiencing the event. Stratification allows the comparison to account for specified baseline factors rather than treating all participants as if they came from a single homogeneous stratum.
What does the BMFS hazard ratio of 0.97 mean?
Under the reported stratified Cox model, the estimated hazard of the BMFS event for denosumab relative to placebo was represented by an HR of 0.97. Because the registry states that HR values below 1 favor denosumab, the point estimate lies slightly in that direction. It does not mean a 3% absolute reduction in bone metastasis or death.
Why is the confidence interval important?
A point estimate alone does not show how precisely the treatment effect has been estimated. The 95% CI of 0.82–1.14 for BMFS indicates that the estimate is uncertain and that values on both sides of the null hazard ratio of 1 are compatible with the interval under the stated statistical framework.
Why does the P-value not measure effect size?
The BMFS P-value of 0.70 addresses the strength of evidence against the specified null hypothesis under the testing model. It does not tell us whether an effect is large or small, nor does it give the probability that the treatment is ineffective. Effect size is described by the hazard ratio and its confidence interval.
What is the role of stratification?
The analysis notes state that the analyses were stratified by breast cancer therapy/LN status, hormone receptor status, HER-2 status, age, and geographic region. Stratification is useful when these factors are incorporated into the randomized comparison so that the time-to-event test and Cox model account for differences across those prespecified strata.
Why does censoring matter in BMFS?
BMFS is defined from randomization to the first occurrence of bone metastasis or death from any cause. Participants last known to be alive without bone metastasis were censored at their last assessment date, or at the primary analysis data cut-off date, whichever was first. Consequently, the analysis uses both event information and follow-up information from participants who did not experience the event during observed follow-up.
11. Understanding the Primary Endpoint
BMFS combines two clinically distinct events into a single time-to-event outcome: the first occurrence of bone metastasis or death from any cause, whichever comes first. This construction means that a participant who dies without a documented bone metastasis can still experience the BMFS event through death.
Participants who remained alive without bone metastasis were censored according to the registry definition.
This distinction is statistically important. A composite time-to-event endpoint is not equivalent to analyzing bone metastasis alone. The treatment comparison reflects the combined event definition specified in the registry.
12. Reading the Hazard Ratio Correctly
The BMFS hazard ratio of 0.97 is a relative comparison of estimated event hazards under the stratified Cox model. A value of 1 represents equal estimated hazards between the groups.
An HR of 0.97 does not mean that 97% of participants remained event-free, that the absolute event probability was reduced by 3%, or that each participant experienced the same relative change in risk.
The 95% CI of 0.82–1.14 gives the uncertainty interval around the primary hazard-ratio estimate. Its inclusion of 1 is consistent with the reported P-value of 0.70 under the two-sided testing framework.
The registry specifies that a hazard ratio below 1 favors denosumab. The primary point estimate is 0.97, but the confidence interval extends on both sides of 1, so the point estimate should not be interpreted independently of its uncertainty.
13. Statistical Interpretation of the Secondary Analyses
| Endpoint | Point estimate relative to 1 | 95% CI includes 1? | Statistical reading |
|---|---|---|---|
| BMFS | 0.97 | Yes | Primary estimate is close to the null; CI spans both directions. |
| DFS | 1.04 | Yes | Estimate is close to the null and slightly above it. |
| Postmenopausal DFS | 1.12 | Yes | Subset estimate is above 1 with a CI spanning 1. |
| Overall Survival | 1.01 | Yes | Estimate is very close to the null. |
| Distant Recurrence-free Survival | 1.06 | Yes | Estimate is close to the null and slightly above it. |
This table is a statistical reading of the reported estimates, not a ranking of endpoints. The estimates should be interpreted in their respective populations, time frames, endpoint definitions, and inferential roles.
14. Stratification and the Cox Model
The registry analysis notes identify five stratification dimensions: breast cancer therapy/LN status, hormone receptor status, HER-2 status, age, and geographic region. The reported Cox model was stratified by these randomization stratification factors.
Why stratify?
Stratification allows the survival comparison to account for prespecified factors used in the trial's randomized design.
What remains common?
The treatment effect is summarized by a hazard ratio comparing denosumab with placebo across the stratified analysis framework.
A stratified Cox model does not mean that separate hazard ratios must be reported for every stratum. Instead, the model can account for different baseline hazards across strata while estimating the treatment comparison within the specified model.
15. Analysis Populations and Their Meaning
| Analysis | Population reported | Why it matters |
|---|---|---|
| Primary BMFS | All randomized participants | The primary efficacy estimate is anchored to randomized treatment assignment. |
| DFS | All randomized participants | Provides a randomized-group time-to-event comparison. |
| Postmenopausal DFS | Randomized participants postmenopausal at enrollment | This is a subset analysis and therefore describes a narrower population than the overall randomized analysis. |
| Overall Survival | All randomized participants | Uses the complete randomized population for the posted analysis. |
| Distant Recurrence-free Survival | All randomized participants | Uses the complete randomized population for the posted analysis. |
The distinction between the overall randomized population and the postmenopausal subset is especially important when interpreting the widths of confidence intervals. A subset analysis addresses a narrower population and should not automatically be treated as interchangeable with the overall trial analysis.
16. Multiplicity and Secondary Endpoints
The registry data identify one primary endpoint and four secondary statistical analyses. All five posted analyses have the hypothesis type superiority, and all use two-sided 95% confidence intervals.
| Endpoint role | Number reported | Interpretation |
|---|---|---|
| Primary | 1 | BMFS is the registered primary endpoint. |
| Secondary | 4 | DFS, postmenopausal DFS, overall survival, and distant recurrence-free survival. |
17. Interim Analysis, Crossover, and Missing Data
The registry-reported D-CARE trial data do not provide a reported interim-analysis procedure, crossover rule, missing-data/imputation method, or Bayesian analysis method. Those design features therefore are not characterized here.
18. Trial Timeline
D-CARE begins
The phase 3 randomized study begins with denosumab and placebo as the two study interventions.
Primary analysis data cut-off
The BMFS primary endpoint and several secondary endpoints use 31 August 2017 as the primary analysis data cut-off date.
Overall survival analysis
The posted overall-survival analysis uses follow-up through the end of study, with median time on study reported as 72.7 (0, 92) and 72.3 (0, 92) months.
19. Limitations and Interpretation Issues
- Primary estimate near the null: the BMFS HR is 0.97, and its 95% CI of 0.82–1.14 spans the null value of 1.
- Hazard-ratio interpretation: an HR is a relative time-to-event measure and should not be converted directly into an absolute risk difference or an individual probability.
- Proportional-hazards framework: the hazard ratio comes from a Cox proportional-hazards model. The ClinicalTrials.gov record does not report a separate assessment of the proportional-hazards assumption.
- Subset analysis: the postmenopausal DFS analysis uses a narrower analysis population and should not be interpreted as equivalent to the all-randomized analysis.
- Secondary endpoint multiplicity: the ClinicalTrials.gov record identifies four secondary analyses but do not specify a multiplicity-adjustment strategy.
- Censoring: BMFS includes censoring for participants last known to be alive without bone metastasis, as specified in the registry definition. Additional censoring sensitivity analyses are not reported.
- Safety interpretation: serious adverse-event counts are reported as affected/at-risk counts and are not themselves a formal adjusted safety comparison.
- Registry scope: this analysis is limited to the facts and statistical analyses contained in the ClinicalTrials.gov record.
20. Why This Trial Matters Statistically
D-CARE is a useful teaching example because the trial brings several fundamental survival-analysis concepts together in a large randomized phase 3 design. The primary endpoint is explicitly time-to-event, the event definition incorporates both bone metastasis and death, and the registry reports a stratified log-rank test together with a stratified Cox hazard ratio.
| Concept | How it appears in D-CARE |
|---|---|
| Randomization | 4,509 participants were enrolled in a randomized two-arm phase 3 trial. |
| Parallel design | The registry identifies the design model as parallel. |
| Quadruple masking | The registry identifies the masking as quadruple. |
| Time-to-event endpoint | BMFS is the registered primary endpoint. |
| Composite event definition | BMFS ends at first bone metastasis or death from any cause, whichever comes first. |
| Censoring | Participants alive without bone metastasis are censored according to the registered definition. |
| Stratified log-rank test | Used for the primary and posted secondary time-to-event comparisons. |
| Hazard ratio | Used as the effect measure for all five posted statistical analyses. |
| Stratified Cox model | Used to obtain the hazard-ratio estimates, stratified by the randomization factors. |
| Confidence intervals | All posted hazard ratios have two-sided 95% confidence intervals. |
| Superiority testing | The hypothesis type is reported as superiority for all five analyses. |
21. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
22. Related Statistical Calculators
23. Sources
- ClinicalTrials.gov: D-CARE, NCT01077154.
- PubMed: PubMed record for PMID 38979716.
Continue through Clinical Biostats
Explore statistical tutorials and calculators related to survival analysis, hazard ratios, confidence intervals, randomization, and clinical-trial methods.
24. Record Summary
D-CARE is a randomized, parallel-group, quadruple-masked phase 3 trial of denosumab versus placebo in women with high-risk early breast cancer receiving neoadjuvant or adjuvant therapy. The registered primary endpoint, BMFS, is a time-to-event outcome defined by the first occurrence of bone metastasis or death from any cause. Its posted analysis used all randomized participants, a stratified log-rank test, and a stratified Cox proportional-hazards model. The reported primary hazard ratio was 0.97, with a two-sided 95% CI of 0.82–1.14 and a P-value of 0.70.
The four posted secondary analyses used the same general survival-analysis framework. The reported hazard ratios were 1.04 for DFS, 1.12 for DFS in the postmenopausal subset, 1.01 for overall survival, and 1.06 for distant recurrence-free survival. Their respective two-sided 95% confidence intervals all included 1, with P-values of 0.57, 0.26, 0.94, and 0.41.
Statistically, the most important lesson is that the primary BMFS result should be read as a complete time-to-event estimate: the hazard ratio, its confidence interval, the testing result, the endpoint definition, the analysis population, and the stratified survival model all contribute to its interpretation. A P-value alone does not describe the size or precision of an effect.