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
TAX316 was a randomized phase 3 parallel-group trial in breast cancer comparing a docetaxel-containing regimen, TAC, with FAC. The registered primary endpoint was disease-free survival, defined using relapse, second primary cancer, or death from any cause.
| Feature | TAX316 |
|---|---|
| Phase | Phase 3 |
| Condition | Breast Cancer |
| Brief title | Docetaxel in Node Positive Adjuvant Breast Cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1491 |
| Primary endpoint type | Time-to-event |
| Primary analysis | Log-rank test with hazard ratio as the effect measure |
| Hypothesis type | Superiority |
| Trial status | Completed |
2. Clinical Question
The central statistical question was whether TAC, the docetaxel-containing regimen, differed from FAC with respect to disease-free survival in patients with node-positive breast cancer.
Population
Participants with node-positive breast cancer enrolled in the TAX316 randomized phase 3 trial.
Intervention
TAC, consisting of Docetaxel, Doxorubicin, and Cyclophosphamide.
Comparator
FAC, consisting of 5-fluorouracil, Doxorubicin, and Cyclophosphamide.
Primary question
Does TAC produce a different disease-free survival experience than FAC, under a prespecified superiority framework?
3. Trial Design
Docetaxel-containing regimen
- Docetaxel
- Doxorubicin
- Cyclophosphamide
5-fluorouracil-containing regimen
- 5-fluorouracil
- Doxorubicin
- Cyclophosphamide
The randomized parallel-group structure is important statistically because treatment assignment establishes the framework for comparing subsequent time-to-event outcomes. The absence of masking is a separate design characteristic; it does not change the mathematical definition of the hazard ratio or log-rank test, but it is part of the overall interpretation of the trial.
4. Registered Primary Endpoint
| Endpoint | Time frame | Definition | Analysis |
|---|---|---|---|
| Number of Participants With Disease-Free Survival Events | up to 10 year follow-up | Disease-Free Survival (DFS)- are defined as local, regional or metastatic relapse or the date of second primary cancer or death from any cause whichever occurs first. | Log Rank; hazard ratio from Cox Proportional Hazard |
This endpoint combines several clinically important first events into a single time-to-event outcome. A participant is counted as having a DFS event when the first qualifying event occurs, rather than requiring all components of the composite definition to occur.
5. Statistical Methodology
Log-rank test
The registry reports a log-rank test for the primary DFS comparison. The log-rank test compares the observed pattern of events between treatment groups across follow-up rather than reducing the outcome to a single fixed-time proportion.
The method is appropriate for a randomized comparison in which participants can experience the endpoint at different follow-up times and some observations can be censored.
Cox proportional-hazards model
The effect measure reported for the primary analysis is a Cox proportional-hazards estimate expressed as a hazard ratio. The hazard ratio summarizes the relative instantaneous event rate between TAC and FAC under the fitted model.
An HR below 1 indicates a lower estimated instantaneous event rate for TAC relative to FAC under the model. It is not a direct probability, risk difference, or percentage of participants experiencing an event.
Superiority framework
The registered hypothesis type is superiority. The primary comparison therefore asks whether the treatment groups differ in the specified direction of the statistical hypothesis rather than whether TAC is merely close enough to FAC to satisfy a non-inferiority margin.
Planned power
The study was originally designed to have 90% power to detect a 26% risk reduction of relapsing for TAC compared to FAC, corresponding to a planned hazard ratio of 0.74.
Power is a design quantity, not a description of the observed treatment effect. The planned HR of 0.74 represents the effect size the study was designed to detect; it should not be confused with the subsequently reported DFS HR of 0.795.
6. Primary Results: Disease-Free Survival
The registry reports a formal statistical analysis for the primary endpoint, comparing TAC with FAC over up to 10 year follow-up.
Hazard ratio for disease-free survival events
95% CI: 0.679–0.932 · P = 0.0043
Two-sided confidence interval · Log-rank test · Cox proportional-hazards effect measure
| Primary endpoint | TAC vs FAC |
|---|---|
| Outcome | Number of Participants With Disease-Free Survival Events |
| Time frame | up to 10 year follow-up |
| Analysis method | Log Rank |
| Effect measure | Cox Proportional Hazard |
| Hazard ratio | 0.795 |
| 95% CI | 0.679–0.932 |
| P-value | 0.0043 |
| Hypothesis | Superiority |
The estimated hazard ratio of 0.795 means that, under the Cox model, the estimated instantaneous rate of experiencing a DFS event in the TAC group was approximately 20.5% lower than in the FAC group. This is a relative time-to-event measure, not a statement that 20.5% of participants avoided relapse or death.
The estimate does not mean that every participant had the same reduction in event risk. It also does not give the probability that an individual participant will remain disease-free at a particular future time.
The 95% confidence interval, 0.679 to 0.932, describes statistical uncertainty around the estimated hazard ratio. It is narrower than an interval extending across substantially different treatment-effect directions, but it still represents uncertainty about the magnitude of the relative treatment effect.
The P-value of 0.0043 addresses the evidence against the null hypothesis under the specified statistical test. It does not measure the size or clinical importance of the treatment effect. Effect magnitude is better described using the hazard ratio and its confidence interval.
Because the effect measure comes from a Cox proportional-hazards model, interpretation of a single hazard ratio relies on the model's proportional-hazards framework. The hazard ratio should therefore not be interpreted as a fixed risk ratio at every individual time point.
Reading the primary result together
The three main statistical quantities answer different questions:
HR = 0.795
Describes the estimated relative event rate between TAC and FAC under the Cox model.
95% CI = 0.679–0.932
Describes uncertainty around the estimated hazard ratio.
P = 0.0043
Quantifies evidence against the null hypothesis under the reported statistical test; it is not an effect-size measure.
Superiority
Defines the hypothesis framework used for the primary comparison rather than a non-inferiority or equivalence framework.
7. Secondary Results: Overall Survival
The registry also reports a secondary time-to-event endpoint, Number of Participants With Overall Survival Events, over up to 10 year follow-up. The comparison was TAC versus FAC using a log-rank test and a Cox proportional-hazards effect measure.
Hazard ratio for overall survival events
95% CI: 0.613–0.898 · P = 0.0020
Two-sided confidence interval · Log-rank test · Cox proportional-hazards effect measure
| Secondary endpoint | TAC vs FAC |
|---|---|
| Outcome | Number of Participants With Overall Survival Events |
| Time frame | up to 10 year follow-up |
| Analysis method | Log Rank |
| Effect measure | Cox Proportional Hazard |
| Hazard ratio | 0.742 |
| 95% CI | 0.613–0.898 |
| P-value | 0.0020 |
| Hypothesis | Superiority |
The overall-survival hazard ratio of 0.742 corresponds to an approximately 25.8% lower estimated hazard for TAC relative to FAC under the Cox model.
As with the DFS result, this is not equivalent to saying that 25.8% of participants avoided death or that an individual patient's probability of survival increased by 25.8%. The hazard ratio is a relative time-to-event measure.
The 95% confidence interval of 0.613–0.898 describes uncertainty around the estimated relative hazard. The P-value of 0.0020 provides evidence against the null hypothesis under the reported log-rank framework, but it does not quantify the magnitude of benefit.
The overall-survival result should also be distinguished conceptually from DFS. DFS can be triggered by several types of first events, whereas overall survival concerns the time-to-event outcome represented by the registry's overall-survival analysis. The two endpoints therefore capture different aspects of the clinical course.
8. Primary and Secondary Results Side by Side
| Endpoint | Role | HR | 95% CI | P-value | Method |
|---|---|---|---|---|---|
| Disease-Free Survival Events | Primary | 0.795 | 0.679–0.932 | 0.0043 | Log-rank; Cox proportional hazard |
| Overall Survival Events | Secondary | 0.742 | 0.613–0.898 | 0.0020 | Log-rank; Cox proportional hazard |
The two estimates are directionally consistent: both are below 1, and both confidence intervals lie below 1. The statistical interpretation remains endpoint-specific, however. A treatment effect on DFS should not automatically be described as the same effect on overall survival, even when both estimates point in the same direction.
9. Safety: Serious Adverse Events by Arm
The registry reports serious adverse events by treatment arm as affected participants over the number at risk.
| Arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| TAC (Docetaxel) | 267 | 744 |
| FAC (5-fluorouracil) | 67 | 736 |
The registry's safety summary is presented separately from the efficacy analyses. The denominators are part of the reported safety information and should be retained when interpreting the event counts. An efficacy hazard ratio cannot be used to summarize serious adverse-event differences because safety and time-to-event efficacy endpoints answer different questions.
10. Statistical Methods Explained
Why was a log-rank test used?
The primary endpoint is a time-to-event outcome measured over follow-up rather than a simple binary outcome at one fixed time. The log-rank test compares the event experience of the randomized treatment groups across the observed follow-up while incorporating the timing of events.
This matters because two groups could have the same eventual number of events but different timing of those events. A time-to-event comparison retains information about when the events occurred.
What does a hazard ratio of 0.795 mean?
A hazard ratio of 0.795 means that the estimated instantaneous rate of a DFS event for TAC was approximately 79.5% of that for FAC under the Cox proportional-hazards model. Expressed as a relative reduction, this corresponds to approximately 20.5% lower estimated hazard.
It does not mean that 20.5% of patients were prevented from experiencing a DFS event. It also does not mean that every patient experienced exactly the same relative reduction.
Why is the confidence interval important?
The point estimate alone does not communicate how precisely the treatment effect has been estimated. The 95% confidence interval of 0.679–0.932 provides a range describing statistical uncertainty around the primary hazard-ratio estimate.
A confidence interval is therefore complementary to the P-value. The P-value addresses evidence against the null hypothesis, while the confidence interval gives a more informative description of the plausible magnitude of the estimated effect under the statistical framework.
Why does the P-value not measure treatment-effect size?
The primary P-value of 0.0043 describes the strength of evidence against the null hypothesis under the reported log-rank test. It does not tell the reader whether the treatment effect is large, small, or clinically important on its own.
For that reason, a statistically interpretable trial result should present the P-value together with the hazard ratio and confidence interval rather than treating the P-value as a measure of benefit.
Why was the trial designed around a hazard ratio of 0.74?
The registry states that the study was originally designed with 90% power to detect a 26% risk reduction of relapsing for TAC compared with FAC, corresponding to a planned hazard ratio of 0.74.
This is a design assumption. It describes the effect size used when planning the study and is not the observed result. The observed primary hazard ratio was 0.795, which should be interpreted independently of the original design target.
Why is the Cox proportional-hazards assumption important?
The reported effect measure is a Cox proportional-hazards hazard ratio. This model summarizes the relative hazard under a proportional-hazards framework. If the relative hazards vary substantially over time, a single hazard ratio may not fully describe the treatment difference at every point during follow-up.
The hazard ratio is therefore best interpreted together with the time-to-event framework, confidence interval, follow-up period, and the underlying survival curves when those curves are available.
11. Understanding Censoring in a Time-to-Event Trial
Time-to-event analysis must distinguish an observed event from a participant who reaches the end of available follow-up without an observed event. The latter contributes information up to the time at which follow-up ends or observation is otherwise censored.
This is why survival methods such as Kaplan-Meier estimation, the log-rank test, and Cox regression are useful for endpoints followed over time.
The registry identifies the primary endpoint as time-to-event and reports a log-rank analysis with a Cox proportional-hazards effect measure. The page therefore interprets the reported results within that survival-analysis framework rather than converting them into a fixed-time binary analysis.
12. Disease-Free Survival as a Composite Time-to-Event Endpoint
The registry definition of DFS combines local, regional or metastatic relapse, second primary cancer, and death from any cause, with whichever qualifying event occurs first defining the DFS event.
Multiple possible events
The endpoint can be triggered by several clinically distinct events.
First event matters
The definition uses whichever qualifying event occurs first.
Time matters
The analysis evaluates when the first qualifying event occurs rather than simply counting participants with events.
Interpretation is composite
A DFS effect represents the combined endpoint definition rather than a result for relapse alone.
This distinction is important when communicating the result. A statement that TAC reduced the hazard of DFS events is more precise than describing the hazard ratio as a direct measure of relapse alone, because the registered DFS definition includes several possible first events.
13. Statistical Interpretation of the Primary Endpoint
The primary HR of 0.795 indicates a lower estimated instantaneous rate of DFS events for TAC relative to FAC under the Cox model. The corresponding simple relative-hazard interpretation is approximately a 20.5% lower estimated hazard.
The 95% CI of 0.679–0.932 indicates uncertainty around the point estimate. The interval provides substantially more information about the estimated effect than the point estimate alone.
The two-sided P-value of 0.0043 is the reported statistical evidence from the primary log-rank comparison. It should not be interpreted as the probability that the treatment effect is real, nor as a percentage reduction in risk.
A hazard ratio does not directly provide an absolute risk difference, a number needed to treat, a median event-free time, or an individual patient's probability of remaining disease-free at a particular time. Those quantities require corresponding underlying survival information.
14. Comparing the Planned and Observed Hazard Ratios
| Quantity | Value | Meaning |
|---|---|---|
| Planned hazard ratio | 0.74 | Effect size the study was designed to detect |
| Observed DFS hazard ratio | 0.795 | Estimated treatment effect from the reported primary analysis |
| Planned power | 90% | Design probability associated with detecting the prespecified effect under the planning assumptions |
| Planned risk reduction | 26% | Risk-reduction assumption corresponding to the planned HR of 0.74 |
These quantities should not be conflated. Statistical power is determined before the outcome data are observed, while the reported hazard ratio is estimated from the trial data. A study can be adequately powered for its design target without the observed estimate being identical to that target.
15. Multiplicity and Hypothesis Interpretation
The registry identifies one registered primary endpoint and separately identifies overall survival as a secondary outcome. This hierarchy matters when interpreting the reported P-values.
| Endpoint | Role | Reported HR | Reported P-value |
|---|---|---|---|
| Disease-Free Survival Events | Primary | 0.795 | 0.0043 |
| Overall Survival Events | Secondary | 0.742 | 0.0020 |
A P-value from a secondary endpoint should not automatically be interpreted as if it had the same confirmatory status as the primary endpoint. The endpoint hierarchy provides essential context for understanding what each analysis was designed to establish.
The available registry information does not report an additional multiplicity-adjustment procedure for the two posted analyses. The safest interpretation is therefore to preserve the distinction between the primary DFS result and the secondary overall-survival result rather than treating both P-values as interchangeable confirmatory evidence.
16. What the Hazard Ratio Does — and Does Not — Mean
An HR of 0.795 means that the estimated instantaneous DFS-event rate under TAC was approximately 79.5% of that under FAC under the fitted Cox model.
The hazard ratio does not mean that an individual participant has a 79.5% probability of remaining disease-free, nor does it mean that 20.5% of participants are guaranteed to avoid a DFS event.
The hazard ratio is not the same mathematical quantity as a risk ratio at a specified follow-up time. It summarizes relative event rates within a time-to-event model.
Absolute treatment benefit requires absolute event probabilities or survival estimates at specified times. Those quantities are distinct from the hazard ratio.
17. Limitations
- Composite endpoint: DFS combines relapse, second primary cancer, and death from any cause. The primary hazard ratio therefore represents the registered composite endpoint rather than relapse alone.
- Hazard-ratio interpretation: the reported effect measure comes from a Cox proportional-hazards model, so interpretation depends on the proportional-hazards framework.
- Relative rather than absolute measure: the hazard ratio does not directly communicate an absolute difference in event-free probability at a particular time.
- Endpoint hierarchy: overall survival is reported as a secondary endpoint and should not simply be treated as statistically interchangeable with the primary DFS endpoint.
- Safety comparison: the registry provides serious-adverse-event counts by arm but does not provide a comparative inferential analysis for those counts in the posted statistical analyses.
- Unreported details: the registry information summarized here does not provide stratification factors, crossover procedures, Bayesian methods, a non-inferiority margin, or a missing-data/imputation strategy. These design features should not be inferred from the reported hazard ratios.
- Follow-up: the registered endpoint time frame extends to up to 10 year follow-up, so interpretation should be understood within that specified follow-up horizon.
18. Why This Trial Matters Statistically
TAX316 is a useful teaching example because it illustrates how a randomized oncology trial can translate a clinically defined endpoint into a formal time-to-event comparison.
| Concept | How it appears in TAX316 |
|---|---|
| Randomization | Participants were randomized to two parallel treatment arms. |
| Time-to-event endpoint | The primary endpoint was disease-free survival over up to 10 year follow-up. |
| Composite endpoint | DFS included relapse, second primary cancer, or death from any cause, whichever occurred first. |
| Log-rank test | Used for the reported TAC versus FAC time-to-event comparison. |
| Hazard ratio | Used to quantify the relative event rate under the Cox proportional-hazards model. |
| Confidence interval | Quantifies uncertainty around the hazard-ratio estimate. |
| P-value | Provides evidence against the null hypothesis under the reported statistical test. |
| Superiority | The registered hypothesis type for the primary comparison. |
| Design power | The study was originally designed with 90% power to detect a 26% risk reduction, corresponding to HR 0.74. |
| Safety analysis | Serious adverse events are reported descriptively by treatment arm. |
19. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary DFS comparison produced a hazard ratio of 0.795 with a 95% confidence interval of 0.679–0.932 and a two-sided P-value of 0.0043 using the reported log-rank framework.
Clinical interpretation
The primary endpoint is a composite time-to-event measure encompassing relapse, second primary cancer, and death from any cause. The reported hazard ratio therefore describes the relative DFS-event experience represented by that definition.
The secondary overall-survival analysis produced a hazard ratio of 0.742 with a 95% confidence interval of 0.613–0.898 and a P-value of 0.0020. This provides a separate time-to-event assessment and should be interpreted as an overall-survival result rather than as another measurement of DFS.
20. Statistical Methods in Practical Terms
Start with the endpoint
DFS is defined by the first qualifying event, so the statistical analysis must account for both event occurrence and event timing.
Use the randomized groups
The treatment comparison is TAC versus FAC, preserving the randomized parallel-group structure.
Compare event patterns
The log-rank test evaluates the time-to-event experience between the treatment groups.
Quantify the relative effect
The Cox proportional-hazards model expresses the comparison as a hazard ratio with a confidence interval.
This sequence is useful when reading clinical-trial results generally: first determine exactly what constitutes an event, then identify the time scale and censoring structure, then identify the statistical comparison, and finally interpret the effect estimate and its uncertainty.
21. Trial Timeline
Study initiated
The registry lists June 1997 as the trial start.
Primary completion
The registry lists January 2010 as the primary completion date.
Trial completed
The registry identifies TAX316 as a completed phase 3 randomized trial.
Primary endpoint horizon
The registered DFS endpoint uses a time frame of up to 10 year follow-up.
22. Results in Context
The primary DFS result and secondary overall-survival result point in the same direction, but they remain distinct statistical analyses.
| Question | DFS | Overall survival |
|---|---|---|
| Endpoint role | Primary | Secondary |
| Time frame | up to 10 year follow-up | up to 10 year follow-up |
| Analysis | Log-rank | Log-rank |
| Effect measure | Cox proportional hazard | Cox proportional hazard |
| HR | 0.795 | 0.742 |
| 95% CI | 0.679–0.932 | 0.613–0.898 |
| P-value | 0.0043 | 0.0020 |
The important statistical lesson is that consistency between endpoints can be informative without making the endpoints interchangeable. DFS and overall survival represent different event definitions, and the primary-versus-secondary hierarchy remains part of the interpretation.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: NCT00688740.
- PubMed: PMID 15930421.
- PubMed: PMID 40207789.
- PubMed: PMID 23246022.
Continue through the Clinical Biostats knowledge graph
Connect this trial's endpoints and statistical methods to deeper tutorials, calculators, and additional clinical-trial analyses.
26. Record Summary
TAX316 is a clear example of a randomized oncology trial in which a clinically defined composite endpoint is analyzed as a time-to-event outcome. The primary DFS analysis used a log-rank test and reported a Cox proportional-hazards estimate of 0.795 with a 95% confidence interval of 0.679–0.932 and a two-sided P-value of 0.0043. The secondary overall-survival analysis reported a hazard ratio of 0.742, with a 95% confidence interval of 0.613–0.898 and a P-value of 0.0020.
The most important statistical reading is not the P-value alone. The hazard ratio describes the relative time-to-event effect, the confidence interval describes uncertainty around that estimate, and the endpoint definition determines exactly what an event represents. Together, these elements provide a more complete statistical interpretation of the randomized comparison than any single summary number.