← Clinical Trials
Hormone-Responsive Breast Cancer Phase 3 Time-to-Event Analysis NCT00066690

SOFT: Complete Statistical Analysis of Ovarian Function Suppression in Hormone-Responsive Breast Cancer

An independent statistical review of the randomized phase 3 SOFT trial comparing tamoxifen alone with tamoxifen plus ovarian function suppression and exemestane plus ovarian function suppression in premenopausal women with hormone-responsive breast cancer.

Trial status: Completed  ·  Enrollment: 3066  ·  Primary completion: 2014-05
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical results on this page are restricted to the ClinicalTrials.gov record.

1. Trial at a Glance

SOFT was a randomized, parallel-group phase 3 trial evaluating three treatment strategies in premenopausal women with hormone-responsive breast cancer: tamoxifen, tamoxifen plus ovarian function suppression, and exemestane plus ovarian function suppression.

3066
Enrollment
Phase 3 trial
3
Treatment arms
Randomized allocation
0.83
DFS HR
Tamoxifen vs T+OFS
0.68
DFS HR
Tamoxifen vs E+OFS
FeatureSOFT
PhasePhase 3
ConditionEstrogen receptor positive breast cancer; progesterone receptor positive tumor; recurrent breast carcinoma; Stage IA, IB, IIA, IIB, and IIIA breast cancer
DesignRandomized, parallel-group, unmasked
AllocationRandomized
Primary endpointDisease-free Survival
Primary endpoint typeTime-to-event
Primary hypothesisSuperiority
Lead sponsorETOP IBCSG Partners Foundation
Sponsor typeNetwork
ClinicalTrials.govNCT00066690
StatusCompleted

2. Clinical Question

The central statistical question was whether treatment strategies incorporating ovarian function suppression differed from tamoxifen alone with respect to disease-free survival, the registered primary time-to-event endpoint.

Population

Premenopausal women with the hormone-responsive breast cancer conditions represented in the registry, including estrogen receptor positive breast cancer and progesterone receptor positive tumor.

Intervention

The trial included two ovarian-function-suppression strategies: tamoxifen plus ovarian function suppression and exemestane plus ovarian function suppression.

Comparator

Tamoxifen alone served as the reference group in the reported hazard-ratio estimations.

Primary question

Do the randomized treatment comparisons demonstrate a difference in disease-free survival under the registered superiority framework?

3. Trial Design

01
Randomize3066 participants
02
3 armsTamoxifen, T+OFS, E+OFS
03
FollowTime-to-event outcomes
04
AnalyzeITT survival comparisons
05
EstimateHazard ratios and 95% CIs
Allocation
Randomized allocation was used to assign participants to the three parallel treatment arms.
Masking
The registry describes the trial as having no masking.
Primary purpose
Treatment.
Analysis framework
The posted primary and secondary efficacy analyses use an intention-to-treat population and time-to-event methodology.
ARM 1

Tamoxifen

  • Tamoxifen
ARM 2

Tamoxifen + OFS

  • Tamoxifen
  • Ovarian function suppression
ARM 3

Exemestane + OFS

  • Exemestane
  • Ovarian function suppression

The registry intervention list also includes laboratory biomarker analysis, oophorectomy, quality-of-life assessment, radiation therapy, and triptorelin. The ClinicalTrials.gov record does not provide enough detail to reconstruct how each of these components was assigned or administered, so this page does not infer additional treatment schedules.

4. Endpoints

EndpointTime frameTypeRegistered definition / analysis context
Disease-free Survival 5-year estimates, reported at a median follow-up of 67 months. Time-to-event Estimated percentage of patients alive and disease-free at 5 years from randomization, where disease-free survival is defined as the time from randomization to the first appearance of one of the following: invasive breast cancer recurrence at local, regional, or distant site, invasive contralateral breast cancer, second (non-breast) invasive cancer, or death without cancer event; or censored at da
Breast Cancer-free Interval 5-year estimates, reported at a median follow-up of 67 months. Time-to-event Secondary endpoint analyzed in the intention-to-treat population.
Distant Recurrence-free Interval 5-year estimates, reported at a median follow-up of 67 months. Time-to-event Secondary endpoint analyzed in the intention-to-treat population.
Overall Survival 8-year estimates, reported at a median follow-up of 8 years Time-to-event Secondary endpoint analyzed in the intention-to-treat population.
Registry-definition note: The ClinicalTrials.gov extract truncates the final censoring phrase in the disease-free-survival definition at “censored at da”. This page therefore does not complete or reconstruct that wording from another source.

5. Statistical Methodology

Intention-to-treat analysis

All posted statistical analyses identify the intention-to-treat population as the analysis population. In an ITT analysis, participants are evaluated according to their randomized treatment assignment rather than being reclassified according to treatment actually received. This preserves the comparison created by randomization.

Log-rank test

Where the registry reports a formal method, the comparison is described as a log-rank test. The log-rank test is designed for comparing survival distributions between groups while accounting for differing follow-up times and right censoring.

Conceptual survival comparison
H0: S1(t) = S0(t)   across the analyzed follow-up

The log-rank framework compares the observed and expected event patterns between groups over time. It is not a comparison of only the final number of events.

Hazard ratio

The principal effect measure reported for the SOFT analyses is the hazard ratio (HR). Tamoxifen was the reference group for the disease-free-survival and breast-cancer-free-interval comparisons. For the distant recurrence-free interval and overall survival analyses, the registry notes identify the reference as “T”; in context, the comparison is reported as Tamoxifen versus the specified treatment group, but this page does not expand that registry abbreviation beyond the ClinicalTrials.gov record.

Conceptual interpretation
HR < 1  →  lower estimated instantaneous event rate in the first-named comparison group relative to its reference

A hazard ratio is a relative time-to-event measure. It is not an absolute risk difference, a probability of benefit, or a statement that every participant experiences the same proportional change in risk.

Confidence intervals

The posted effect estimates are accompanied by two-sided 95% confidence intervals. The interval expresses statistical uncertainty around the estimated hazard ratio under the analysis framework. It should not be interpreted as the range in which individual patient effects must fall.

Superiority testing

The posted analyses identify the hypothesis type as superiority. This differs conceptually from a non-inferiority design: the statistical objective is to establish evidence of a difference rather than to demonstrate that a treatment is not worse than a prespecified margin.

6. Results: Primary Endpoint

The registered primary endpoint was disease-free survival, with 5-year estimates reported at a median follow-up of 67 months. Two primary statistical comparisons are posted, both based on the intention-to-treat population.

Tamoxifen vs T+OFS

Disease-free survival hazard ratio

0.83

95% CI: 0.66–1.04   ·   P = 0.1

Analysis population: Intention-to-treat  ·  Method: Log-rank  ·  Hypothesis: Superiority

Clinical Biostats interpretation

The reported HR of 0.83 means the estimated hazard was lower in the T+OFS comparison relative to the tamoxifen reference under the reported time-to-event analysis. Expressed as a simple relative-hazard interpretation, an HR of 0.83 corresponds to an estimated 17% lower hazard.

The estimate does not mean that 17% fewer participants had an event, that 17% of participants benefited, or that each participant's individual probability of recurrence was reduced by exactly 17%. Those interpretations would require absolute event probabilities or other clinically interpretable measures.

The two-sided 95% CI of 0.66–1.04 describes uncertainty around the estimated hazard ratio. Because the interval includes 1, the reported interval is compatible with no difference as well as with a range of lower-hazard effects under the model and sampling framework.

The reported P = 0.1 is evidence about compatibility with the null hypothesis under the specified testing framework; it is not a measure of the size or clinical importance of the HR. The magnitude of the HR and its confidence interval should therefore be considered separately from the P-value.

Tamoxifen vs E+OFS

Disease-free survival hazard ratio

0.68

95% CI: 0.53–0.86

Analysis population: Intention-to-treat  ·  Hypothesis: Superiority

The registry extract does not report a formal statistical-analysis method or P-value for this comparison.

Clinical Biostats interpretation

The reported HR of 0.68 corresponds to an estimated hazard approximately 32% lower for the E+OFS comparison relative to tamoxifen under the reported hazard-ratio framework.

The 95% CI of 0.53–0.86 indicates uncertainty around the estimated relative hazard. Every value in that interval is a value for the population-level hazard ratio that remains compatible with the stated confidence procedure; it is not a prediction interval for individual patients.

The ClinicalTrials.gov record does not report a P-value or formal method for this specific comparison. It would therefore be inappropriate to manufacture a P-value or attribute a particular testing procedure to the comparison. The evidence available in the registry extract is the reported HR and its confidence interval.

The estimate should also not be read as a fixed proportional reduction in event probability at every time point. A hazard ratio summarizes relative event rates over time and is model-dependent; interpretation can be affected if the proportional-hazards assumption is not appropriate.

Primary endpoint comparison

ComparisonEndpointAnalysis populationMethodHR95% CIP-value
Tamoxifen vs T+OFS Disease-free Survival Intention-to-treat Log Rank 0.83 0.66–1.04 0.1
Tamoxifen vs E+OFS Disease-free Survival Intention-to-treat Not reported 0.68 0.53–0.86 Not reported

7. Secondary Endpoint Results

Breast Cancer-free Interval

ComparisonAnalysis populationMethodHR95% CIP-value
Tamoxifen vs T+OFS Intention-to-treat Log Rank 0.81 0.3–1.03 0.09
Tamoxifen vs E+OFS Intention-to-treat Not reported 0.64 0.49–0.83 Not reported

Both breast-cancer-free-interval analyses use the same 5-year time frame, reported at a median follow-up of 67 months. The T+OFS estimate of 0.81 corresponds to an estimated hazard approximately 19% lower than the tamoxifen reference, while the E+OFS estimate of 0.64 corresponds to an estimated hazard approximately 36% lower. These are relative hazard interpretations, not absolute reductions in the probability of a breast-cancer event.

Distant Recurrence-free Interval

ComparisonAnalysis populationMethodHR95% CIP-value
Tamoxifen vs T+OFS Intention-to-treat Log Rank 0.88 0.66–1.18 0.40
Tamoxifen vs E+OFS Intention-to-treat Not reported 0.71 0.52–0.96 Not reported

For T+OFS, the reported HR of 0.88 corresponds to an estimated hazard approximately 12% lower than the tamoxifen reference. Its 95% CI of 0.66–1.18 includes 1, and the reported P-value is 0.40. For E+OFS, the reported HR is 0.71 with a 95% CI of 0.52–0.96; the registry extract does not report a formal method or P-value for that comparison.

Reference-group terminology: The ClinicalTrials.gov record for the distant recurrence-free interval states “T” as the reference group in its analysis notes. The comparison itself is recorded as Tamoxifen vs T+OFS or Tamoxifen vs E+OFS. This page preserves the comparison labels reported by the registry rather than silently expanding the abbreviated reference notation.

Overall Survival

ComparisonTime frameAnalysis populationMethodHR95% CIP-value
Tamoxifen vs T+OFS 8-year estimates, reported at a median follow-up of 8 years Intention-to-treat Log Rank 0.67 0.48–0.92 0.01
Tamoxifen vs E+OFS 8-year estimates, reported at a median follow-up of 8 years Intention-to-treat Not reported 0.85 0.62–1.15 Not reported

The T+OFS overall-survival estimate of 0.67 corresponds to an estimated hazard approximately 33% lower than the tamoxifen reference. The two-sided 95% CI is 0.48–0.92, and the reported P-value is 0.01. For E+OFS, the reported HR is 0.85 with a 95% CI of 0.62–1.15; no formal method or P-value is reported in the registry-reported extract.

8. Secondary Results: How to Read the Pattern

The posted results contain several distinct endpoint families, and they should not be collapsed into one summary statistic. Disease-free survival includes multiple possible events, breast cancer-free interval focuses on breast-cancer-related events, distant recurrence-free interval focuses on distant recurrence, and overall survival concerns death. Each therefore represents a different clinical event process.

Relative effect

The HR communicates the relative event rate associated with one randomized treatment group compared with its reference. HRs below 1 indicate lower estimated hazard for the first-named comparison under the reported framework.

Absolute effect

The statistical analyses posted on ClinicalTrials.gov do not provide the corresponding absolute survival percentages or event probabilities for these comparisons. The HR therefore cannot be converted into an absolute patient-level risk reduction from the ClinicalTrials.gov record alone.

Precision

The width of each 95% confidence interval reflects statistical uncertainty. For example, the disease-free-survival interval for T+OFS is 0.66–1.04, while the E+OFS interval is 0.53–0.86.

Time frame

The disease-free-survival and breast-cancer-free-interval analyses report 5-year estimates, whereas overall survival is reported using 8-year estimates. These time frames should not be treated as interchangeable.

9. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk.

Treatment armSerious adverse eventsAt risk
Tamoxifen3151007
T+OFS3751007
E+OFS3751001

These figures provide arm-level counts of participants affected by serious adverse events and the corresponding numbers at risk. They should be interpreted as a safety summary rather than as an efficacy comparison. The ClinicalTrials.gov record does not provide event-specific breakdowns, severity categories, exposure durations, treatment-emergent definitions, or formal comparative safety tests, so none are added here.

Statistical interpretation of safety counts

The denominators matter. Reporting only the affected counts would obscure the amount of information contributed by each arm. Here, the registry provides both affected and at-risk counts, allowing the reader to see the scale of the safety population in each arm without assuming that all participants had identical treatment exposure.

These counts also should not be interpreted as evidence for or against an overall treatment strategy without additional information about event definitions, timing, severity, exposure, competing risks, and statistical comparisons. A serious adverse-event count answers a different question from a time-to-event efficacy endpoint.

10. Statistical Methods Explained

Why was intention-to-treat analysis used?

The posted analyses use the intention-to-treat population. This approach retains participants in the groups to which they were randomized, preserving the treatment comparison generated by the randomization process. It is particularly important in a multi-arm randomized trial because excluding or reassigning participants after randomization can introduce selection into the comparison.

What does a hazard ratio of 0.68 mean?

An HR of 0.68 indicates an estimated instantaneous event hazard 32% lower for the first-named treatment comparison relative to its tamoxifen reference, under the reported hazard-ratio framework. It does not mean that 32% of participants avoided an event, nor does it mean that each participant's individual risk was reduced by 32%.

Why is the confidence interval important?

A point estimate alone does not communicate how precisely the treatment effect has been estimated. The 95% CI places the reported HR in the context of sampling uncertainty. For the E+OFS disease-free-survival comparison, the interval is 0.53–0.86; for T+OFS, it is 0.66–1.04. Those intervals convey different levels of compatibility with values around the null value of 1.

Why is a P-value not an effect-size measure?

A P-value depends on the observed data and the statistical model under a null hypothesis, as well as the amount of information available. It does not tell the reader whether an HR of 0.68 is clinically large or small. Effect magnitude is described by the HR, while the confidence interval communicates its statistical precision.

Why use a log-rank test for a time-to-event endpoint?

The log-rank test is suited to comparing groups when the timing of events and censoring matter. A simple comparison of event proportions at a single administrative time would discard information about when events occurred and how long participants were followed.

Why should the 5-year and 8-year analyses be kept separate?

The disease-free-survival and breast-cancer-free-interval results use 5-year estimates at a median follow-up of 67 months, whereas overall survival is reported using 8-year estimates at a median follow-up of 8 years. Different follow-up horizons represent different amounts of accumulated information and should not be combined as though they came from the same analysis time point.

What does the proportional-hazards assumption have to do with an HR?

The hazard ratio is a model-based summary of relative event rates over time. When the proportional-hazards assumption is not appropriate, a single HR can become less descriptive of the full survival experience. The ClinicalTrials.gov record reports HRs but do not provide a proportional-hazards diagnostic, so this page does not claim that the assumption was verified.

11. Confidence Intervals and P-values in Context

The SOFT registry results illustrate why an HR, confidence interval, and P-value should be read together rather than separately.

ComparisonHR95% CIP-valueWhat the ClinicalTrials.gov record supports
DFS: Tamoxifen vs T+OFS0.830.66–1.040.1Estimated lower hazard; interval includes 1; formal log-rank analysis reported.
DFS: Tamoxifen vs E+OFS0.680.53–0.86Not reportedEstimated lower hazard; interval does not include 1; formal method not reported in registry-reported extract.
BCFI: Tamoxifen vs T+OFS0.810.3–1.030.09Estimated lower hazard; interval includes 1; formal log-rank analysis reported.
BCFI: Tamoxifen vs E+OFS0.640.49–0.83Not reportedEstimated lower hazard; interval does not include 1; formal method not reported in registry-reported extract.
Distant recurrence-free interval: Tamoxifen vs T+OFS0.880.66–1.180.40Estimated lower hazard; interval includes 1; formal log-rank analysis reported.
Distant recurrence-free interval: Tamoxifen vs E+OFS0.710.52–0.96Not reportedEstimated lower hazard; interval does not include 1; formal method not reported in registry-reported extract.
OS: Tamoxifen vs T+OFS0.670.48–0.920.01Estimated lower hazard; interval does not include 1; formal log-rank analysis reported.
OS: Tamoxifen vs E+OFS0.850.62–1.15Not reportedEstimated lower hazard; interval includes 1; formal method not reported in registry-reported extract.

The table also highlights an important registry-data limitation: formal methods and P-values are not reported for every posted comparison. A complete statistical analysis should distinguish between a result that has an explicitly documented test and a result for which only an effect estimate and confidence interval are available.

12. Multiplicity and Multiple Comparisons

The trial has one registered primary endpoint but the posted statistical analyses include multiple treatment comparisons and several secondary endpoints. The ClinicalTrials.gov record identifies the hypothesis type as superiority, but they do not provide an alpha-allocation strategy, multiplicity-adjustment procedure, endpoint hierarchy beyond the designation of the primary endpoint, or familywise-error-control plan.

What is documented

Disease-free survival is the single registered primary endpoint. Multiple comparisons involving T+OFS and E+OFS are posted, along with secondary endpoints.

What is not documented here

The registry-reported extract does not state how type I error was allocated across treatment comparisons or secondary endpoints.

This distinction matters when interpreting nominal P-values. A P-value reported for one comparison does not, by itself, establish how that result should be interpreted within the complete family of statistical tests. Because the ClinicalTrials.gov record does not specify the multiplicity procedure, this page does not infer one.

13. Randomization and the Three-Arm Structure

Randomization is the core design feature that makes the treatment comparisons interpretable as randomized comparisons rather than observational associations. SOFT used three parallel treatment arms, allowing two ovarian-function-suppression strategies to be compared with tamoxifen.

Reference strategy
Tamoxifen was the reference group in the reported hazard-ratio estimation for the primary disease-free-survival comparisons.
Comparison 1
Tamoxifen vs T+OFS, with an HR of 0.83 for disease-free survival.
Comparison 2
Tamoxifen vs E+OFS, with an HR of 0.68 for disease-free survival.
Analysis population
The posted efficacy analyses use intention-to-treat analysis.

A three-arm design also means that the statistical interpretation cannot simply be reduced to a binary treatment-versus-placebo framework. Each randomized comparison has its own effect estimate, confidence interval, and, where reported, P-value.

14. What the Hazard Ratios Do — and Do Not — Mean

Disease-free survival

The HR of 0.83 for Tamoxifen vs T+OFS represents an estimated lower hazard in the T+OFS comparison relative to tamoxifen. The HR of 0.68 for Tamoxifen vs E+OFS represents a larger estimated relative reduction in hazard under the same interpretation.

These estimates do not establish that the treatment effects are identical at every point in time, nor do they establish that the absolute probability of an event is reduced by the same percentage. They summarize a time-to-event comparison using the reported hazard-ratio framework.

Confidence intervals

The 95% CIs quantify uncertainty around each reported HR. For example, the T+OFS disease-free-survival estimate has a CI of 0.66–1.04, while the E+OFS estimate has a CI of 0.53–0.86. The intervals should be interpreted as statistical uncertainty around the population-level estimate, not as a range of outcomes for individual patients.

P-values

For the T+OFS disease-free-survival comparison, the reported P-value is 0.1; for the T+OFS overall-survival comparison, it is 0.01. The registry extract does not provide P-values for the E+OFS comparisons. A P-value does not quantify treatment magnitude and should not be used as a substitute for the HR or its confidence interval.

15. Time-to-Event Endpoints and Censoring

SOFT's primary endpoint is explicitly classified as a time-to-event outcome. This means that both whether an event occurred and when it occurred contribute to the statistical analysis.

Why time matters
Observed information = event timing + follow-up time + censoring information

Participants who have not experienced the event by the end of available follow-up can contribute information without being treated as though they experienced the event.

Censoring is therefore a central part of survival analysis rather than a nuisance that can simply be discarded. The Kaplan-Meier and log-rank frameworks are designed to incorporate censored observations. The registry extract does not, however, provide detailed censoring counts or censoring-pattern diagnostics, so no additional assumptions about censoring can be verified from the ClinicalTrials.gov record.

16. Primary Endpoint vs Secondary Endpoints

EndpointRoleTime frameComparisons posted
Disease-free SurvivalPrimary5-year estimates, reported at a median follow-up of 67 months.Tamoxifen vs T+OFS; Tamoxifen vs E+OFS
Breast Cancer-free IntervalSecondary5-year estimates, reported at a median follow-up of 67 months.Tamoxifen vs T+OFS; Tamoxifen vs E+OFS
Distant Recurrence-free IntervalSecondary5-year estimates, reported at a median follow-up of 67 months.Tamoxifen vs T+OFS; Tamoxifen vs E+OFS
Overall SurvivalSecondary8-year estimates, reported at a median follow-up of 8 yearsTamoxifen vs T+OFS; Tamoxifen vs E+OFS

This hierarchy matters because a primary endpoint is the endpoint designated in advance as the principal question of the trial. Secondary endpoints can provide additional information, but they should not automatically be treated as equivalent to the primary endpoint for confirmatory inference.

17. Trial Timeline

2003-12-17

Trial start

The registry lists December 17, 2003 as the study start date.

2014-05

Primary completion

The registry lists May 2014 as the primary completion date.

5-year endpoint analysis

Disease-free survival and related endpoints

The posted disease-free-survival, breast-cancer-free-interval, and distant recurrence-free-interval analyses use 5-year estimates reported at a median follow-up of 67 months.

8-year endpoint analysis

Overall survival

The posted overall-survival analyses use 8-year estimates reported at a median follow-up of 8 years.

18. Limitations

19. Why This Trial Matters Statistically

SOFT is a useful statistical teaching case because it combines a randomized three-arm design with a time-to-event primary endpoint and multiple related efficacy outcomes. The analysis demonstrates why treatment effects need to be read at several levels: the randomized comparison, the endpoint definition, the time horizon, the hazard ratio, the confidence interval, and the formal hypothesis-testing information.

ConceptHow it appears in SOFT
RandomizationThree parallel randomized treatment arms.
Intention-to-treat analysisPosted efficacy analyses use the intention-to-treat population.
Time-to-event endpointDisease-free survival is the registered primary endpoint.
Log-rank testReported for several of the posted treatment comparisons.
Hazard ratioUsed as the principal effect measure across the posted statistical analyses.
Confidence intervalTwo-sided 95% CIs accompany all eight posted statistical analyses.
SuperiorityThe posted hypothesis type is superiority.
Multiple comparisonsTwo ovarian-function-suppression strategies are compared with tamoxifen across multiple endpoints.
Different follow-up horizonsPrimary and several secondary analyses use 5-year estimates, while overall survival uses 8-year estimates.
Safety by armSerious adverse-event counts are reported for all three treatment groups.

20. Statistical Interpretation vs Clinical Interpretation

Statistical interpretation

The posted disease-free-survival analyses produce HR estimates of 0.83 for Tamoxifen vs T+OFS and 0.68 for Tamoxifen vs E+OFS. The corresponding 95% CIs are 0.66–1.04 and 0.53–0.86. The registry reports P = 0.1 for the first comparison and no P-value for the second.

Clinical interpretation

The endpoints represent different clinical event processes, so their results should be considered separately. The ClinicalTrials.gov record supports description of the relative time-to-event estimates but do not provide enough information to quantify absolute treatment benefit for an individual patient.

A useful distinction is that statistical evidence describes what was estimated under the trial's analysis framework, whereas clinical interpretation asks what those estimates mean in the context of patient outcomes. The latter requires information beyond a single hazard ratio, including absolute risks, duration of follow-up, adverse events, treatment burden, and patient characteristics. The registry extract does not contain all of those components.

21. Reading the SOFT Results as a Statistical Story

The primary disease-free-survival comparison with T+OFS has an HR below 1, but its 95% CI extends above 1 and its reported P-value is 0.1. The E+OFS comparison has an HR of 0.68 with a 95% CI of 0.53–0.86, but the registry extract does not report a P-value or formal method for that comparison.

The secondary endpoints add additional dimensions rather than simply repeating the primary endpoint. Breast cancer-free interval shows HRs of 0.81 and 0.64 for the two comparisons, while distant recurrence-free interval shows HRs of 0.88 and 0.71. Overall survival, reported at the longer 8-year time frame, shows HRs of 0.67 and 0.85.

These results should not be collapsed into a single average effect. Each HR belongs to a particular endpoint, treatment comparison, follow-up horizon, and statistical analysis. A statistically careful reading preserves those distinctions.

Key statistical lesson: The most informative unit of interpretation is not simply “the trial had a positive or negative result.” It is the combination of endpoint + comparison + analysis population + effect estimate + confidence interval + testing information + follow-up time.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue through the Clinical Biostats statistical pathway

Use the related tutorials and calculators to explore the survival-analysis concepts represented in the SOFT trial, including hazard ratios, confidence intervals, randomization, and time-to-event methods.

25. Record Summary

SOFT provides a clear example of how a randomized three-arm phase 3 trial can generate several related but distinct time-to-event analyses. The registered primary endpoint was disease-free survival, with 5-year estimates reported at a median follow-up of 67 months. The posted primary analyses produced hazard ratios of 0.83 for Tamoxifen vs T+OFS and 0.68 for Tamoxifen vs E+OFS. Secondary analyses extended the statistical picture to breast cancer-free interval, distant recurrence-free interval, and overall survival, with overall survival reported using 8-year estimates at a median follow-up of 8 years.

The most important statistical lesson is that an HR is only one component of an evidence statement. Its interpretation depends on the endpoint, randomized comparison, analysis population, confidence interval, testing procedure, follow-up horizon, and assumptions underlying the time-to-event model. The SOFT registry record also demonstrates why analysts should distinguish clearly between methods that are explicitly reported and methods that would merely be plausible choices for a particular endpoint.

Clinical Biostats methodology: A trial-results page should not merely repeat an outcome table. The goal is to reconstruct the statistical structure of the trial while separating reported evidence from educational interpretation and preserving uncertainty where the registry does not provide enough information to support a stronger claim.