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
EURAMOS-1 was a randomized, open-label, parallel phase 3 treatment trial in patients with localized or metastatic osteosarcoma. The registry reports four treatment arms and 1334 enrolled participants, with event-free survival as the primary endpoint.
| Feature | EURAMOS-1 |
|---|---|
| Phase | Phase 3 |
| Conditions | Localized Osteosarcoma; Metastatic Osteosarcoma |
| Design | Randomized, parallel |
| Allocation | Randomized |
| Masking | None |
| Primary purpose | Treatment |
| Primary endpoint | Event-free Survival (EFS) |
| Primary endpoint type | Time-to-event |
| Enrollment | 1334 |
| Trial status | Completed |
| Lead sponsor | Children's Oncology Group |
| Start | November 14, 2005 |
| Primary completion | January 31, 2015 |
2. Clinical Question
The central statistical question was whether randomized treatment strategies produced different event-free survival in patients with localized or metastatic osteosarcoma. The primary endpoint was evaluated separately for the good-response and poor-response randomized treatment groups represented by the reported comparisons.
Population
Patients with localized or metastatic osteosarcoma enrolled in the EURAMOS-1 phase 3 trial.
Interventions
The registry lists cisplatin, doxorubicin hydrochloride, etoposide, ifosfamide, methotrexate, peginterferon alfa-2b, therapeutic conventional surgery, quality-of-life assessment, and questionnaire administration.
Randomized comparisons
The posted primary analyses compare MAP-GR with MAPifn and MAP-PR with MAPIE.
Primary question
Do the randomized treatment strategies differ in time to the first event defining event-free survival?
3. Trial Design
The registry classifies EURAMOS-1 as randomized, with a parallel design and no masking. The primary purpose is treatment. The four-arm structure is important because the primary efficacy analyses are not presented as one four-group omnibus comparison; instead, the posted analyses identify two randomized pairwise comparisons.
MAP-GR vs MAPifn
- MAP-GR is one randomized treatment strategy.
- MAPifn is the comparator randomized treatment strategy.
- The posted primary analysis evaluates event-free survival.
- The reported effect measure is a hazard ratio.
MAP-PR vs MAPIE
- MAP-PR is one randomized treatment strategy.
- MAPIE is the comparator randomized treatment strategy.
- The posted primary analyses evaluate event-free survival.
- Both a hazard-ratio analysis and an RMST difference are reported.
4. Trial Timeline
Trial start
The registry records November 14, 2005 as the study start date.
Primary completion
The registry records January 31, 2015 as the primary completion date.
Registry status
The trial is listed as completed, with results posted and five statistical analyses posted.
5. Endpoints
| Endpoint | Registry definition / time frame | Statistical role |
|---|---|---|
| Event-free Survival (EFS) | EFS is defined as time from randomisation to the first of: death, detection of local recurrence or metastasis, progression of metastatic disease, or detection of a secondary malignancy. From date of randomization to date of the event. | Primary endpoint; time-to-event |
| Percentage of Patients With Overall Survival | From date of randomization to date of death. | Secondary endpoint; time-to-event |
The registry states that EFS will be assessed using the logrank test and expressed using hazard ratios with appropriate confidence intervals. Follow-up per participant will be assessed for up to 10 years, and the registry specifies that the 3-year EFS is provided as a summary.
The posted statistical analyses contain three primary EFS analyses and two secondary overall-survival analyses. The registry reports effect estimates, 95% two-sided confidence intervals, and p-values for all five analyses.
6. Statistical Methodology
Log-rank test
The primary EFS comparisons use the log-rank test. This is a standard method for comparing time-to-event distributions between randomized groups when participants can be followed for different lengths of time and some participants have not experienced the event by the end of observation.
The key feature is that the analysis uses the ordering of event times and the risk sets at those times rather than reducing follow-up to a single binary outcome. This is particularly important for EFS because the registry defines an event as the first occurrence of several clinically relevant outcomes.
Hazard ratio
The principal effect measure for the posted EFS and overall-survival analyses is the hazard ratio. A hazard ratio compares the estimated instantaneous event rates between two randomized groups over the analyzed follow-up.
For example, an HR of 0.83 corresponds to an estimated hazard approximately 17% lower in the first-listed group than in the comparator, because 1 − 0.83 = 0.17. This is a relative hazard interpretation, not a statement that 17% of patients avoided an event.
Mean difference and restricted mean survival time
For the MAP-PR vs MAPIE EFS comparison, the registry also reports a difference in RMST. Restricted mean survival time summarizes the expected event-free time over a prespecified restriction horizon. A difference in RMST is therefore expressed on an absolute time scale, unlike a hazard ratio.
The registry specifically notes that the secondary RMST analysis was performed in the poor-response group because of evidence of non-proportional hazards. This is an important methodological point: when the relative hazard is not adequately represented by a constant proportional relationship over time, an RMST comparison can provide a complementary summary that does not rely on the proportional-hazards assumption in the same way.
Superiority framework
The posted analyses identify the hypothesis type as superiority. Therefore, the relevant question is whether the randomized groups differ in the direction and magnitude supported by the observed estimate, rather than whether one treatment is merely no worse than another by a prespecified non-inferiority margin.
7. Primary Results: Event-Free Survival
The registry posts three primary EFS analyses. Two are hazard-ratio analyses from the randomized pairwise comparisons, while the third is an RMST analysis in the poor-response comparison.
MAP-GR vs MAPifn
Event-free survival hazard ratio
95% CI: 0.61–1.12 · P = 0.214
Two-sided 95% confidence interval; log-rank analysis; superiority hypothesis.
The estimated hazard ratio of 0.83 indicates a lower estimated instantaneous rate of an EFS event in MAP-GR relative to MAPifn, corresponding to approximately a 17% lower estimated hazard under the hazard-ratio interpretation.
The estimate does not mean that 17% more patients were event-free, nor does it mean that every patient experienced a 17% reduction in event risk. It is a relative time-to-event measure.
The 95% confidence interval of 0.61–1.12 is relatively wide and includes 1.00. Thus, the observed estimate is compatible with a range of treatment effects that includes little or no difference as well as a lower event hazard for MAP-GR. The interval describes uncertainty around the estimated relative hazard; it does not describe the range of individual patient outcomes.
The p-value of 0.214 quantifies the evidence against the relevant null hypothesis under the specified testing framework. It does not measure the size or clinical importance of the estimated effect. Because the registry identifies the analysis as a superiority test, the interpretation should focus on the estimated effect together with its confidence interval rather than treating the p-value as a standalone measure of benefit.
MAP-PR vs MAPIE
Event-free survival hazard ratio
95% CI: 0.78–1.23 · P = 0.86
Two-sided 95% confidence interval; log-rank analysis; superiority hypothesis.
The estimated hazard ratio of 0.98 is very close to 1.00. Under the hazard-ratio interpretation, the estimated instantaneous rate of an EFS event was approximately 2% lower in MAP-PR than in MAPIE.
The confidence interval of 0.78–1.23 crosses 1.00 and therefore includes both a possible reduction and a possible increase in the event hazard. The interval is more informative than the point estimate alone because it shows the uncertainty surrounding the observed comparison.
The p-value of 0.86 indicates little evidence against the null hypothesis in this superiority comparison. It should not be interpreted as proving that the two strategies are exactly equivalent. In a superiority trial, failure to demonstrate superiority is not the same statistical claim as establishing equivalence or non-inferiority.
MAP-PR vs MAPIE: RMST analysis
Difference in restricted mean survival time
95% CI: −3.3 to 4.9 · P = 0.69
Mean Difference (Final Values); two-sided 95% confidence interval.
The estimated RMST difference of 0.8 represents an estimated absolute difference in restricted mean event-free time, with MAP-PR minus MAPIE according to the reported comparison. The positive point estimate therefore favors MAP-PR on this particular time-scale summary.
The 95% confidence interval extends from −3.3 to 4.9, spanning both negative and positive values. The observed estimate is therefore compatible with a range of possible differences in either direction.
The registry identifies this as a secondary RMST analysis because of evidence of non-proportional hazards in the poor-response group. That makes the RMST analysis particularly useful as a complementary summary: it describes average event-free time over the restriction period rather than relying on a single proportional-hazards parameter.
The p-value of 0.69 does not provide evidence of a statistically detectable RMST difference under the reported superiority framework. It is not a measure of how clinically meaningful a difference of 0.8 is.
8. Primary EFS Results in One View
| Comparison | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| MAP-GR vs MAPifn | Log-rank | Hazard ratio | 0.83 | 0.61–1.12 | 0.214 |
| MAP-PR vs MAPIE | Log-rank | Hazard ratio | 0.98 | 0.78–1.23 | 0.86 |
| MAP-PR vs MAPIE | Difference in RMST | Mean difference | 0.8 | −3.3 to 4.9 | 0.69 |
The three primary analyses point to an important statistical distinction. The MAP-GR vs MAPifn hazard-ratio estimate is below 1, whereas the MAP-PR vs MAPIE estimate is very close to 1. In the poor-response comparison, the RMST analysis provides an additional absolute-time summary because the registry reports evidence of non-proportional hazards.
9. Secondary Results: Overall Survival
The registry also posts two secondary analyses of Percentage of Patients With Overall Survival, defined from randomization to death.
MAP-GR vs MAPifn
Overall survival hazard ratio
95% CI: 0.69–1.33 · P = 0.804
Two-sided 95% confidence interval; log-rank analysis; superiority hypothesis.
An overall-survival hazard ratio of 0.96 indicates an estimated instantaneous rate of death approximately 4% lower in MAP-GR than MAPifn under the reported hazard-ratio comparison.
The 95% confidence interval of 0.69–1.33 is broad enough to include both a potentially meaningful reduction and a potentially higher hazard. Because it crosses 1.00, the estimate does not provide a precise indication of a directional difference in overall survival.
The p-value of 0.804 provides little evidence against the null hypothesis in the reported superiority analysis. As with EFS, this does not prove identical survival distributions; it indicates that the observed data do not provide strong evidence for superiority under the stated analysis.
MAP-PR vs MAPIE
Overall survival hazard ratio
95% CI: 0.81–1.39 · P = 0.674
Two-sided 95% confidence interval; log-rank analysis; superiority hypothesis.
An overall-survival hazard ratio of 1.06 indicates an estimated instantaneous rate of death approximately 6% higher in MAP-PR than MAPIE under the reported comparison.
The 95% confidence interval of 0.81–1.39 crosses 1.00, so the estimate is compatible with both a lower and a higher mortality hazard. The confidence interval should be considered alongside the point estimate rather than treating 1.06 as an exact measure of a treatment difference.
The p-value of 0.674 does not provide evidence of a statistically detectable superiority difference in overall survival for this comparison. It does not establish equivalence between MAP-PR and MAPIE.
10. Overall Survival Results in One View
| Comparison | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| MAP-GR vs MAPifn | Log-rank | Hazard ratio | 0.96 | 0.69–1.33 | 0.804 |
| MAP-PR vs MAPIE | Log-rank | Hazard ratio | 1.06 | 0.81–1.39 | 0.674 |
Overall survival is a different endpoint from EFS because it counts death as the event and does not include the other EFS events specified by the registry. Consequently, an EFS result and an OS result should not be expected to have the same hazard ratio or statistical interpretation.
11. Safety: Serious Adverse Events by Arm
The registry reports serious adverse events by randomized treatment arm as affected participants divided by participants at risk.
| Arm | Affected / At risk | Reported proportion |
|---|---|---|
| MAP-GR | 49 / 359 | 49/359 |
| MAPifn | 76 / 357 | 76/357 |
| MAP-PR | 36 / 310 | 36/310 |
| MAPIE | 43 / 308 | 43/308 |
The affected and at-risk counts are reported separately from the efficacy analyses. They should therefore not be substituted for the randomized sample sizes or treated as denominators for the EFS or overall-survival hazard ratios.
The visual proportions above are shown only to make the affected-versus-at-risk counts easier to compare; the registry's reported counts remain the primary quantities. A serious-adverse-event frequency is not directly comparable with a time-to-event hazard ratio because the statistical questions and follow-up structures differ.
12. Statistical Methods Explained
Why was a log-rank test used for EFS?
EFS is a time-to-event endpoint. Patients can experience an event at different times, and some may remain event-free when their observation ends. The log-rank test is designed to compare the survival experience of randomized groups across their observed event times while accounting for these differing follow-up times.
What does an EFS hazard ratio of 0.83 mean?
The estimate of 0.83 means that the modeled instantaneous event hazard for MAP-GR relative to MAPifn was estimated at 83% of the comparator hazard. Expressed as a relative reduction, this corresponds to approximately 17%. It does not mean a 17-percentage-point improvement in EFS or that each participant's probability of an event fell by exactly 17%.
Why is the confidence interval more informative than the point estimate alone?
A hazard ratio is an estimate rather than a known population parameter. The interval around it describes statistical uncertainty. For MAP-GR vs MAPifn, the interval of 0.61–1.12 demonstrates that the observed estimate is compatible with a range of possible underlying relative hazards. Looking only at 0.83 would conceal that uncertainty.
Why was RMST examined for MAP-PR vs MAPIE?
The registry notes evidence of non-proportional hazards in the poor-response group. A standard hazard-ratio summary is most naturally interpreted when the proportional-hazards assumption is reasonable. RMST provides a complementary measure by comparing average event-free time over a specified restriction period rather than requiring the treatment effect to be represented by one constant hazard ratio.
Why is a p-value not an effect size?
A p-value describes the compatibility of the observed data with a specified null hypothesis under the statistical model and testing procedure. It depends on both the magnitude of the observed difference and the amount of information available. The hazard ratio or RMST difference describes the estimated effect itself, while the confidence interval describes its uncertainty.
Does a nonsignificant superiority test prove the treatments are equivalent?
No. The EURAMOS-1 analyses are identified as superiority analyses. A nonsignificant superiority test means that the analysis did not provide sufficient evidence of superiority under the specified framework. Equivalence and non-inferiority require different hypotheses and, in the latter case, a prespecified margin.
Why should EFS and overall survival not be treated as the same endpoint?
EFS includes the first occurrence of death, local recurrence or metastasis, progression of metastatic disease, or a secondary malignancy. Overall survival uses death as the event. Because the event definitions differ, the two endpoints answer different clinical questions even though both use time-to-event methodology.
13. Interpreting the Hazard Ratios Together
| Comparison | EFS HR | OS HR | Interpretive pattern |
|---|---|---|---|
| MAP-GR vs MAPifn | 0.83 (95% CI 0.61–1.12) | 0.96 (95% CI 0.69–1.33) | EFS estimate below 1; OS estimate close to 1; both confidence intervals include 1 |
| MAP-PR vs MAPIE | 0.98 (95% CI 0.78–1.23) | 1.06 (95% CI 0.81–1.39) | EFS estimate close to 1; OS estimate slightly above 1; both confidence intervals include 1 |
This side-by-side presentation illustrates why it is important not to describe a trial using a single p-value or a single effect estimate. The estimated direction and magnitude vary by randomized comparison and endpoint. Confidence intervals also show that the estimates are uncertain.
For MAP-GR vs MAPifn, the EFS point estimate favors MAP-GR, but the 95% confidence interval includes 1.00. The corresponding OS estimate is closer to 1.00 and likewise has a confidence interval crossing 1.00.
For MAP-PR vs MAPIE, the EFS hazard ratio is 0.98, while the OS hazard ratio is 1.06. Neither confidence interval excludes 1.00. The additional RMST analysis gives an estimated difference of 0.8 with a 95% confidence interval from −3.3 to 4.9.
These results should be interpreted within each randomized comparison rather than averaging or informally combining the reported estimates.
14. Randomization and Causal Interpretation
Randomization is the central design feature supporting causal comparison. When participants are randomly assigned, treatment assignment is not chosen by investigators or participants based on observed prognosis in the way it can be in an observational study. This provides the foundation for comparing outcomes between randomized groups.
EURAMOS-1 is explicitly classified as randomized, but the registry also identifies several important cohort-level features that affect interpretation of the available efficacy population.
Good-response cohort
The registry states that 128 of 357 MAPifn patients completed the planned protocol treatment.
Poor-response cohort
The registry states that 42% of registered patients were not randomized in the poor-response group.
Additional registered patients
Further patients were registered and joined a treatment cohort for follow-up but declined randomization or were not eligible.
Interpretive boundary
The registry states that data for these non-randomized patients are not presented in the reported randomized analyses.
The distinction between registration and randomization is statistically important. The causal interpretation of a randomized comparison is anchored to the participants who entered the randomized comparison. Registered patients who were not randomized do not automatically belong to the same estimand.
15. Analysis Population and Treatment Completion
The registry reports a specific treatment-completion issue in the good-response comparison: among the 357 patients randomized to MAPifn, 128 completed the planned protocol treatment.
This illustrates why treatment received and treatment assigned are different statistical concepts. A randomized efficacy analysis generally preserves the original treatment assignment as the comparison variable. Treatment completion, discontinuation, or other deviations can affect observed outcomes without changing the original randomized assignment.
16. Missing Data, Censoring, and Time-to-Event Interpretation
Because EFS and overall survival are time-to-event endpoints, participants do not all need to experience an event for the analysis to use their available follow-up. A participant who has not experienced the relevant event by the end of observation contributes information up to the time at which follow-up ends.
The registry defines EFS from the date of randomization to the date of the event and specifies several possible event types. This means the analysis is fundamentally concerned with when the first qualifying event occurs, rather than simply whether a patient ever experiences any event during an arbitrary observation window.
The appropriate interpretation of a hazard ratio therefore depends on the follow-up process, censoring rules, event definitions, and model assumptions. A hazard ratio is not interchangeable with a simple difference in percentages.
The event definition determines what counts as failure; the follow-up process determines how long each participant contributes information.
17. Non-Proportional Hazards and RMST
The MAP-PR vs MAPIE primary EFS analysis is particularly useful statistically because the registry explicitly notes evidence of non-proportional hazards. Under proportional hazards, the relative hazard between groups is assumed to remain reasonably stable over time. When that assumption is not appropriate, one hazard ratio may not adequately describe the treatment contrast across the entire follow-up period.
The posted RMST analysis provides an alternative summary. Rather than asking for one relative hazard, RMST asks how much event-free time is accumulated on average up to a prespecified restriction time. The difference of 0.8 therefore has an absolute-time interpretation.
Hazard ratio
Relative comparison of instantaneous event hazards; particularly useful when a proportional-hazards summary is appropriate.
RMST difference
Absolute difference in restricted mean event-free time; useful as a complementary summary when hazards are not proportional.
The two approaches are not competing descriptions of the same number. They summarize the treatment comparison on different scales and answer related but distinct statistical questions.
18. Multiplicity and Multiple Statistical Analyses
The registry posts five statistical analyses: three for the primary EFS endpoint and two for the secondary overall-survival endpoint. The EFS analyses comprise two hazard-ratio comparisons and one RMST analysis. The overall-survival analyses comprise one hazard-ratio comparison for each randomized pair.
| Analysis family | Number posted | Role |
|---|---|---|
| Primary EFS | 3 | Two log-rank hazard-ratio analyses plus one RMST analysis |
| Secondary overall survival | 2 | Log-rank hazard-ratio analyses |
| Total statistical analyses | 5 | All posted registry analyses |
When several hypotheses or comparisons are evaluated, the distinction between a nominal p-value and a multiplicity-controlled confirmatory conclusion becomes important. The registry identifies all posted analyses as superiority analyses, but the information available here does not specify a multiplicity-adjustment strategy or alpha-allocation scheme.
Consequently, each p-value should be read in the context of its specific endpoint and randomized comparison rather than assuming that all five analyses constitute one simple, independently interpretable hypothesis test.
19. What the Confidence Intervals Say
The HR estimate is 0.83, with a 95% CI of 0.61–1.12. The interval crosses 1.00, so the plausible range represented by the interval includes no relative hazard difference as well as lower and higher event hazards.
The HR estimate is 0.98, with a 95% CI of 0.78–1.23. The estimate is close to the null value, and the interval includes values on both sides of 1.00.
The mean difference is 0.8, with a 95% CI of −3.3 to 4.9. Because the interval includes zero, it is compatible with an RMST difference in either direction.
The confidence interval should be considered a core part of each result rather than an optional supplement to the p-value. It communicates both the direction of the estimate and the statistical uncertainty surrounding it.
20. What the P-values Do — and Do Not — Mean
The posted p-values range from 0.214 to 0.86 for the three primary EFS analyses and from 0.674 to 0.804 for the two secondary overall-survival analyses.
A p-value is not the probability that the null hypothesis is true. Nor is it the probability that a treatment has no effect. It is a quantity calculated under a specified statistical model and null hypothesis and describes how compatible the observed result is with that null framework.
For clinical interpretation, the effect estimate and confidence interval should be read first. The p-value then provides the corresponding hypothesis-testing context.
21. Limitations
- Four-arm structure: the primary results are reported as two pairwise randomized comparisons rather than one single treatment effect across all four arms.
- Non-randomized participants: the registry states that 42% of registered patients in the poor-response group were not randomized and that additional registered patients declined randomization or were not eligible. Their data are not presented in the randomized analyses described here.
- Treatment completion: among the 357 MAPifn patients in the good-response group, 128 completed the planned protocol treatment. Completion and randomized assignment are distinct quantities.
- Non-proportional hazards: the registry specifically identifies evidence of non-proportional hazards in the poor-response group, motivating the additional RMST analysis.
- Confidence-interval width: the reported intervals span substantial ranges of possible effects, particularly for the MAP-GR vs MAPifn EFS comparison.
- Multiple analyses: five statistical analyses are posted across the primary and secondary endpoints. The available record does not specify an alpha-allocation or multiplicity-adjustment strategy for interpreting all five estimates jointly.
- Endpoint distinction: EFS and overall survival are different time-to-event endpoints and should not be interpreted as interchangeable measures.
- Safety denominator: serious adverse-event counts are reported using affected/at-risk counts by arm and should not be substituted for the efficacy analysis population.
22. Why This Trial Matters Statistically
EURAMOS-1 is a useful statistical teaching case because its posted analyses demonstrate several important principles in clinical-trial methodology within one randomized phase 3 study.
| Concept | How it appears in EURAMOS-1 |
|---|---|
| Randomization | The registry classifies the study as randomized. |
| Parallel design | The study is classified as a parallel design with four arms. |
| Time-to-event endpoint | EFS is the primary endpoint, defined from randomization to the first qualifying event. |
| Log-rank test | Posted primary and secondary survival analyses use the log-rank method. |
| Hazard ratio | Hazard ratios are the principal reported effect measure for EFS and overall survival. |
| Confidence interval | All posted effect estimates include two-sided 95% confidence intervals. |
| RMST | A difference in RMST is reported for MAP-PR vs MAPIE EFS because of evidence of non-proportional hazards. |
| Superiority testing | The posted analyses are identified as superiority hypotheses. |
| Analysis populations | The registry distinguishes randomized patients from additional registered but non-randomized patients. |
| Safety analysis | Serious adverse events are reported separately by randomized treatment arm. |
The most important lesson is that a clinical trial result is not adequately summarized by a single p-value. The design, endpoint definition, effect measure, confidence interval, analysis population, model assumptions, and structure of the randomized comparisons all contribute to the statistical interpretation.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: EURAMOS-1, NCT00134030.
- PubMed record: PMID 35246418.
- PubMed record: PMID 31401903.
- PubMed record: PMID 27569442.
- PubMed record: PMID 25421877.
Continue through the Clinical Biostats knowledge graph
Connect clinical-trial endpoints to deeper statistical tutorials and practical analysis tools.
26. Record Summary
EURAMOS-1 provides a particularly instructive example of randomized time-to-event analysis in a four-arm phase 3 oncology trial. The registry identifies event-free survival as the primary endpoint and posts three primary statistical analyses: a log-rank hazard-ratio comparison for MAP-GR versus MAPifn, a log-rank hazard-ratio comparison for MAP-PR versus MAPIE, and an RMST difference for MAP-PR versus MAPIE in the setting of evidence of non-proportional hazards.
The MAP-GR versus MAPifn EFS analysis produced an HR of 0.83 with a 95% CI of 0.61–1.12 and P = 0.214. The MAP-PR versus MAPIE EFS analysis produced an HR of 0.98 with a 95% CI of 0.78–1.23 and P = 0.86. The corresponding RMST analysis produced a mean difference of 0.8, with a 95% CI of −3.3 to 4.9 and P = 0.69.
The secondary overall-survival analyses produced HRs of 0.96 for MAP-GR versus MAPifn and 1.06 for MAP-PR versus MAPIE, with both confidence intervals crossing 1.00. The statistical interpretation therefore depends on more than whether a p-value crosses a conventional threshold: the randomized comparison, endpoint definition, effect measure, confidence interval, and proportional-hazards considerations all matter.