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
PORTEC-2 was a randomized phase 3 trial evaluating external-beam radiation therapy versus vaginal brachytherapy after surgery for stage I endometrial cancer. The study used a parallel-group design with two radiation-treatment arms and enrolled 427 participants.
| Feature | PORTEC-2 |
|---|---|
| Phase | Phase 3 |
| Condition | Endometrial Cancer |
| Population | Stage I endometrial cancer after surgery |
| Design | Randomized, parallel-group |
| Allocation | Randomized |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 427 |
| Interventions | External Beam Radiation Therapy; Vaginal Brachytherapy |
| ClinicalTrials.gov | NCT00376844 |
| Lead sponsor | Leiden University |
| Sponsor type | OTHER |
2. Clinical Question
The central clinical question is whether external-beam radiation therapy or vaginal brachytherapy should be compared as postoperative radiation treatments for patients with stage I endometrial cancer. The registered primary endpoint is vaginal relapse at 5 years, including both total vaginal relapse and vaginal relapse as the first failure.
Population
Patients with stage I endometrial cancer after surgery.
Intervention
External Beam Radiation Therapy.
Comparator
Vaginal Brachytherapy.
Primary question
How do the two postoperative radiation approaches compare with respect to vaginal relapse at 5 years?
3. Trial Design
External-Beam Radiation Therapy
- External Beam Radiation Therapy
- Radiation intervention
- Compared directly with vaginal brachytherapy
Vaginal Brachytherapy
- Vaginal Brachytherapy
- Radiation intervention
- Compared directly with external-beam radiation therapy
Why the randomized parallel design matters
Randomization creates the framework for comparing the two radiation strategies while reducing systematic differences in patient characteristics that could otherwise distort the treatment comparison. The parallel structure means each participant belongs to one randomized treatment pathway rather than receiving both interventions as part of a crossover sequence.
The absence of masking is also relevant. A radiation-treatment comparison can be difficult to mask because the treatment procedures themselves differ. Without masking, outcome assessment and other aspects of trial conduct can potentially be influenced by knowledge of treatment assignment, depending on the endpoint and how it is assessed.
4. Trial Timeline
Trial start
The registered study start date was 2002-05.
Primary completion
The registered primary completion date was 2009-05.
Study status
The ClinicalTrials.gov record lists the study status as COMPLETED.
5. Primary Endpoint
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Vaginal relapse | 5 years | Total vaginal relapse and vaginal relapse as first failure |
The primary endpoint is therefore not simply a generic measure of recurrence. The registry defines it through two related descriptions: total vaginal relapse and vaginal relapse as the first failure. The distinction matters statistically because an analysis of all vaginal relapses can answer a different question from an analysis restricted to the first failure experienced by a patient.
Total vaginal relapse versus vaginal relapse as first failure
A total-relapse measure considers vaginal relapse events according to the endpoint definition, whereas a first-failure analysis focuses on whether vaginal relapse is the first recorded failure. These formulations can produce different event counts and different statistical interpretations because subsequent failures are treated differently.
6. Planned Analysis
The registry identifies vaginal relapse, assessed at 5 years, as the primary endpoint. The ClinicalTrials.gov record does not report a posted statistical analysis for this endpoint.
For an endpoint defined as relapse over a fixed follow-up period, a straightforward analysis can compare the proportion of participants experiencing vaginal relapse by 5 years. When the timing of relapse is available, a time-to-event framework can provide additional information by incorporating the timing of events and censoring.
For a first-failure endpoint, time-to-first-failure methods are particularly relevant. If patients can experience other types of failure before vaginal relapse, those alternative failures may act as competing events. In that setting, cumulative-incidence methods can distinguish the probability of vaginal relapse from the probability of other mutually exclusive first failures.
This simple proportion describes an absolute risk at the specified time point. It is different from a hazard ratio, which compares instantaneous event rates over follow-up.
Kaplan-Meier methods are commonly used for time-to-event outcomes when censoring is appropriately handled. When competing first failures are present, cumulative-incidence methods may be more appropriate for estimating the probability of a specific failure type.
7. Statistical Methodology
Binary analysis at 5 years
If the primary question is framed as whether vaginal relapse occurred by 5 years, the outcome can be represented as a binary indicator for each participant: relapse by 5 years versus no recorded relapse by 5 years. The treatment groups can then be compared using an absolute risk difference, risk ratio, odds ratio, or an appropriate hypothesis test for two proportions.
Each effect measure answers a slightly different question. An absolute risk difference describes the difference in percentage points between treatment groups. A risk ratio describes the relative risk. An odds ratio compares odds rather than probabilities. None of these quantities is interchangeable with a hazard ratio.
Time-to-event analysis
If the timing of vaginal relapse is available, time-to-event analysis can use more information than a simple 5-year indicator. Participants who have not experienced vaginal relapse at their last assessment can contribute follow-up information through the point at which they are censored.
Kaplan-Meier estimation is a standard approach for estimating an event-free survival function. A Cox proportional-hazards model can then express the relative event rate through a hazard ratio, provided the model and its assumptions are appropriate.
Competing risks
The phrase "vaginal relapse as first failure" makes competing-risk reasoning particularly important. If another failure occurs first, the participant can no longer experience vaginal relapse as the first failure. Treating such an event as ordinary noninformative censoring can overstate the probability of the vaginal-relapse event.
A competing-risks framework instead recognizes that different first-failure types compete for the same participant. The cumulative incidence function directly estimates the probability of experiencing the specified failure by a given time in the presence of those competing events.
Confidence intervals
A confidence interval quantifies statistical uncertainty around an estimated treatment effect. For a risk difference, the interval describes uncertainty around the difference in absolute risks. For a hazard ratio, it describes uncertainty around the relative event-rate estimate from the fitted survival model.
A hazard ratio is not a probability, not a percentage of patients cured, and not the same quantity as the difference between two 5-year relapse rates.
Intention-to-treat principle
Because PORTEC-2 is randomized, an intention-to-treat analysis would preserve the comparison created by randomization by analyzing participants according to their assigned treatment group. The registry information does not report a specific posted analysis population for the primary endpoint.
8. Statistical Methods Explained
Why is vaginal relapse a useful statistical endpoint?
Vaginal relapse is a clinically specific event that can be defined over a fixed period. Restricting the primary endpoint to a particular type of relapse can focus the treatment comparison on the outcome most directly relevant to the postoperative radiation strategies being compared.
Why distinguish total vaginal relapse from first failure?
A participant can potentially experience more than one type of disease failure over follow-up. "Vaginal relapse" and "vaginal relapse as first failure" therefore describe different event constructions. The first-failure version asks whether vaginal relapse was the first observed failure, while total relapse can encompass vaginal relapse occurring within the broader sequence of disease events.
Why might a 5-year proportion be easier to interpret than a hazard ratio?
A 5-year risk is an absolute quantity tied to a clinically recognizable time point. A hazard ratio summarizes relative event rates over the analyzed follow-up. When the clinical question is explicitly framed at 5 years, the absolute 5-year risk can be especially intuitive.
Why might time-to-event analysis still be preferable?
Participants are not necessarily observed for identical lengths of time. Time-to-event methods can incorporate the timing of relapse and appropriately handle censoring, rather than reducing every participant's follow-up to a single yes-or-no outcome without regard to when the event occurred.
Why do competing risks matter?
If a patient experiences another type of first failure before vaginal relapse, that patient cannot subsequently count as someone whose first failure was vaginal relapse. A competing-risk analysis respects this structure and estimates the probability of the specific failure in the presence of the other possible first failures.
What does a confidence interval tell us?
A confidence interval describes the statistical uncertainty associated with an estimated treatment effect. A narrow interval indicates greater precision than a wide interval, all else equal. It does not describe the range of individual patient responses.
Why does a p-value not measure effect size?
A p-value addresses evidence against a specified null hypothesis under a statistical model. It does not tell us how large the treatment difference is. Effect size should be communicated directly through quantities such as an absolute risk difference, relative risk, odds ratio, or hazard ratio, together with its confidence interval.
9. Understanding the Primary Endpoint Statistically
Absolute risk
The proportion of participants experiencing vaginal relapse by 5 years directly describes the absolute frequency of the event over the registry-defined time frame.
Risk difference
The difference between two 5-year risks expresses the absolute separation between the randomized groups in percentage-point terms.
Risk ratio
The ratio of the two 5-year risks expresses the relative frequency of vaginal relapse by the specified time point.
Hazard ratio
A hazard ratio compares instantaneous event rates over follow-up and should not be substituted for a 5-year risk or risk difference.
These measures are complementary rather than interchangeable. A statistically complete report can present an absolute measure alongside a relative measure so that readers can see both the magnitude of the treatment difference and its scale relative to the comparator.
10. Analysis of Vaginal Relapse as First Failure
The registry's wording specifically includes vaginal relapse "as first failure." This makes the ordering of events part of the endpoint definition. Statistically, that means the analysis is concerned not only with whether vaginal relapse ever occurs, but also with whether it occurs before another qualifying failure.
| Question | Statistical implication |
|---|---|
| Did vaginal relapse occur by 5 years? | A fixed-time risk can summarize the occurrence of vaginal relapse by the specified time point. |
| When did vaginal relapse occur? | A time-to-event analysis can incorporate event timing and censoring. |
| Was vaginal relapse the first failure? | The event ordering must be incorporated into the endpoint definition. |
| Did another failure occur first? | That event can act as a competing first failure when estimating vaginal relapse as first failure. |
This distinction prevents an important analytical error: a method designed for a simple binary outcome should not automatically be assumed to answer a first-failure question when other types of failure can occur before the endpoint of interest.
11. Randomization and Causal Interpretation
The randomized allocation is central to the statistical interpretation of PORTEC-2. If randomization is properly implemented and analyzed according to assignment, differences in outcomes between the two groups can be interpreted within the framework of a randomized treatment comparison rather than merely as an association between treatment choice and outcome.
That does not eliminate every source of uncertainty. Randomization balances characteristics probabilistically rather than guaranteeing identical groups. Sampling variability remains, which is why confidence intervals and hypothesis tests are important components of the analysis.
12. Masking and Outcome Assessment
The registry describes the study as having no masking. Knowledge of treatment assignment can matter more for some outcomes than for others. A relapse endpoint based on clinical or diagnostic assessment can be affected by knowledge of treatment if assessment decisions are not fully standardized.
For this reason, the credibility of an endpoint analysis depends not only on the statistical model but also on how the endpoint was defined, assessed, recorded, and adjudicated. The ClinicalTrials.gov record identifies the primary endpoint but does not provide additional details about adjudication or blinded outcome assessment.
13. What the Registry Reports — and What It Does Not
| Item | ClinicalTrials.gov record |
|---|---|
| Trial identification | PORTEC-2; NCT00376844 |
| Phase | Phase 3 |
| Status | COMPLETED |
| Enrollment | 427 |
| Design | Randomized, parallel |
| Masking | None |
| Primary endpoint | Vaginal relapse at 5 years |
| Primary endpoint definition | Total vaginal relapse and vaginal relapse as first failure |
| Posted statistical analyses | None |
| Posted numerical efficacy results | Not reported in the ClinicalTrials.gov record |
The absence of posted numerical efficacy analyses is important for interpretation. The registry establishes the trial's design and primary endpoint, but it does not provide a numerical estimate, confidence interval, or p-value for the primary endpoint in the information reported here.
14. Limitations
- No posted statistical analysis: the ClinicalTrials.gov record contains no posted statistical analysis for the primary endpoint, so the registry itself does not provide a numerical treatment effect, confidence interval, or p-value.
- Limited endpoint detail: the record identifies vaginal relapse and vaginal relapse as first failure, but does not provide additional operational detail about event adjudication or assessment procedures in the information reported here.
- Competing events: first-failure analyses can require competing-risk methods when alternative failures prevent vaginal relapse from being the first event.
- Censoring: time-to-event methods depend on appropriate handling of participants whose outcome status is not observed through the full follow-up period.
- Unreported analysis population: the registry information does not specify a posted statistical analysis population for the primary endpoint.
- No numerical subgroup information: the registry information reported here does not provide subgroup estimates for the primary endpoint.
- No safety results in the record information reported here: treatment-related safety event counts or rates are not available in the ClinicalTrials.gov information presented for this analysis.
15. Why This Trial Matters Statistically
PORTEC-2 is a useful statistical teaching case because the primary endpoint is more nuanced than a simple overall survival outcome. The trial compares two randomized postoperative radiation strategies while focusing on vaginal relapse at a prespecified 5-year time frame and distinguishing total vaginal relapse from vaginal relapse as the first failure.
| Concept | How it appears in PORTEC-2 |
|---|---|
| Randomization | Participants were randomized to two radiation-treatment strategies. |
| Parallel design | The two interventions were evaluated in separate randomized treatment arms. |
| Fixed follow-up | The primary endpoint is defined at 5 years. |
| Event definition | Vaginal relapse is defined through total relapse and relapse as first failure. |
| Time-to-event reasoning | The timing of relapse can provide information beyond a simple 5-year indicator. |
| Competing risks | Alternative first failures can affect analysis of vaginal relapse as first failure. |
| Absolute effect measures | A 5-year risk difference can describe the absolute separation between treatment groups. |
| Relative effect measures | Risk ratios or hazard ratios can describe relative treatment effects when appropriate. |
| Confidence intervals | Precision should accompany any estimated treatment effect. |
| Masking | The registry identifies the study as unmasked. |
16. Interpreting Treatment Effects Without Overstating Them
If a future analysis reports two 5-year vaginal-relapse risks, their difference would be an absolute risk difference. This answers how many more or fewer relapse events occurred per 100 participants over the specified period, expressed in percentage-point terms.
A risk ratio would describe the relative frequency of vaginal relapse by 5 years, while a hazard ratio would compare event rates over follow-up. Neither should be interpreted as a direct percentage reduction in the number of patients who experience relapse.
Any reported estimate should be accompanied by its confidence interval. The point estimate alone cannot show how precisely the trial estimated the treatment difference.
A p-value, if reported, would address evidence against a specified null hypothesis. It would not replace the treatment-effect estimate or its confidence interval and would not by itself describe clinical magnitude.
17. Fixed-Time Versus Time-to-Event Analysis
The five-year endpoint can be viewed from two complementary perspectives. A fixed-time analysis asks whether the event occurred by 5 years. A time-to-event analysis additionally asks when the event occurred and uses information from participants with different lengths of observed follow-up.
| Approach | Primary question | Key quantity |
|---|---|---|
| Fixed-time risk | What proportion experienced vaginal relapse by 5 years? | 5-year risk |
| Risk difference | How far apart are the two 5-year risks? | Absolute difference |
| Risk ratio | How many times the 5-year risk occurs in one group relative to the other? | Relative risk |
| Kaplan-Meier analysis | How does event-free probability change over time? | Survival/event-free function |
| Cox model | How do event rates compare over follow-up? | Hazard ratio |
| Competing-risk analysis | What is the probability of vaginal relapse as a specific first failure when other failures can occur first? | Cumulative incidence |
These methods are not interchangeable. The appropriate choice depends on the exact endpoint definition, follow-up structure, censoring process, and whether other failure types compete with the event of interest.
18. Primary Endpoint: A Statistical Reading
The phrase "total vaginal relapse and vaginal relapse as first failure" contains two distinct analytical ideas. The first concerns the occurrence of vaginal relapse. The second adds an ordering condition: vaginal relapse must be the first failure.
That ordering condition changes the statistical structure. If another failure happens first, the participant no longer belongs to the group that can experience vaginal relapse as first failure. Consequently, an analysis that ignores competing first failures can answer a different question from the one specified by the endpoint.
19. What a Complete Results Analysis Would Contain
A statistically complete report of the primary endpoint would normally present the number of participants in each randomized group, the number experiencing vaginal relapse, the corresponding 5-year event estimates, and an effect measure with a confidence interval. If a formal hypothesis test were part of the prespecified analysis, its p-value would also be reported.
For a time-to-event or competing-risk analysis, a complete report could additionally show the number of events, censoring information, an estimated event-free curve or cumulative-incidence curve, and the statistical model used to compare the randomized groups.
Effect size
Report an absolute or relative treatment effect rather than relying on a p-value alone.
Uncertainty
Report a confidence interval alongside the point estimate.
Time
Preserve the registered 5-year time frame when interpreting the primary endpoint.
Event structure
Distinguish total vaginal relapse from vaginal relapse as first failure and account for competing events when appropriate.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial provides a randomized comparison of two postoperative radiation strategies with vaginal relapse specified as the primary endpoint at 5 years. The endpoint definition includes both total vaginal relapse and vaginal relapse as first failure.
Clinical interpretation
The trial addresses an important postoperative treatment comparison in stage I endometrial cancer. The clinical meaning of the comparison depends on the observed vaginal-relapse outcomes together with the broader benefits, harms, and treatment burdens reported for the interventions.
The distinction is important: statistical analysis describes the magnitude and uncertainty of the observed treatment comparison, while clinical interpretation considers what those outcomes mean in the context of patient care. Neither should be reduced to a single p-value.
21. Record Summary
PORTEC-2 is a randomized phase 3, parallel-group trial in stage I endometrial cancer after surgery, comparing external-beam radiation therapy with vaginal brachytherapy. The ClinicalTrials.gov record reports an enrollment of 427 participants and identifies vaginal relapse at 5 years as the primary endpoint, defined as total vaginal relapse and vaginal relapse as first failure.
From a statistical perspective, the trial is especially useful for understanding the difference between fixed-time relapse measures and time-to-event methods, and for recognizing why competing risks become important when an endpoint is defined as a specific type of first failure. A rigorous analysis should present an appropriate treatment-effect estimate, its confidence interval, and the statistical method used to account for censoring and competing events where applicable.
22. Sources
- ClinicalTrials.gov: PORTEC-2, NCT00376844.
- PubMed: PORTEC-2 publication record.
- PubMed: PORTEC-2 publication record.
- PubMed: PORTEC-2 publication record.
- PubMed: PORTEC-2 publication record.
- PubMed: PORTEC-2 publication record.
Continue through the Clinical Biostats knowledge graph
Explore clinical trial methodology, survival analysis, and statistical interpretation across the Clinical Biostats library.