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
REVERCE is a randomized, open-label, parallel phase 2 treatment trial with 22 enrolled participants and 2 treatment arms. The primary endpoint is overall survival at a 20-month time frame, with first progression-free survival, second progression-free survival, and sequential treatment progression-free survival reported as secondary time-to-event outcomes.
| Feature | REVERCE |
|---|---|
| Phase | Phase 2 |
| Condition | Locally advanced unresectable or metastatic colorectal adenocarcinoma, including stage III and stage IV disease, with BRAF V600E, KRAS, and NRAS negative status |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 22 |
| Number of arms | 2 |
| Status | Active, not recruiting |
| Lead sponsor | Academic and Community Cancer Research United |
2. Clinical Question
The registry describes a randomized comparison of two treatment strategies in patients with unresectable, locally advanced, or metastatic colorectal cancer who are negative for BRAF V600E, KRAS, and NRAS mutations. The primary statistical question is whether overall survival differs between the two randomized treatment groups at the prespecified 20-month time frame.
Population
Patients with locally advanced unresectable or metastatic colorectal adenocarcinoma, including stage III and stage IV disease, with BRAF V600E, KRAS, and NRAS negative status.
Arm A
Regorafenib.
Arm B
Cetuximab, panitumumab, and irinotecan.
Primary question
Does the regorafenib arm show superior overall survival compared with the cetuximab, panitumumab, and irinotecan arm?
3. Trial Design
Regorafenib
- Regorafenib
Cetuximab, panitumumab, and irinotecan
- Cetuximab
- Panitumumab
- Irinotecan
The registry identifies the allocation as randomized and the design model as parallel. Because masking is listed as none, the trial is open-label rather than blinded. The statistical analysis therefore compares outcomes according to randomized treatment assignment without a masking procedure described in the registry.
4. Trial Population and Analysis Population
The registered condition includes colorectal adenocarcinoma that is locally advanced and unresectable or metastatic, together with specific molecular and stage classifications. The registry's posted statistical analyses state that only eligible patients were included for overall survival, first progression-free survival, and sequential treatment progression-free survival.
| Analysis | Population specified in the registry |
|---|---|
| Overall Survival | Only eligible patients were included in analysis |
| First Progression-free Survival | Only eligible patients were included in analysis |
| Second PFS | Only patients that were eligible and experienced at least one progression were included |
| Sequential Treatment PFS | Only eligible patients were included in analysis |
The restriction for second PFS is statistically important. Unlike overall survival and first PFS, this analysis conditions on having experienced at least one progression. The population entering the second-PFS comparison is therefore a subset of the eligible randomized population rather than the full eligible population.
5. Endpoints
| Endpoint | Role | Time frame | Registry description / analysis |
|---|---|---|---|
| Overall Survival (OS) | Primary | 20 months | The median OS and 95% confidence intervals in each arm will be reported; posted comparison used a one-sided unstratified log-rank test. |
| First Progression-free Survival (PFS) | Secondary | 11 months | Posted comparison used a one-sided unstratified log-rank test. |
| Second PFS | Secondary | 3 months | Analysis included eligible patients who experienced at least one progression; comparison used a one-sided unstratified log-rank test. |
| Sequential Treatment PFS | Secondary | 12 months | Posted comparison used a one-sided unstratified log-rank test. |
The endpoint names and time frames are retained from the registry. The available posted statistical analyses provide P-values for these endpoints, but do not provide treatment-effect estimates or confidence intervals for the comparisons.
6. Statistical Methodology
One-sided unstratified log-rank test
All four posted statistical comparisons use a one-sided unstratified log-rank test. This is a standard method for comparing time-to-event distributions between randomized groups when observations can be censored during follow-up.
The test evaluates the pattern of observed events over time rather than comparing only a single time point. Patients who have not experienced the event by their last assessment can contribute information through their censoring time.
Why the test is unstratified
An unstratified log-rank test compares the randomized groups without incorporating additional stratification factors into the test statistic. The registry specifically identifies the posted method as unstratified, so no adjustment for baseline strata is described in the statistical-analysis record.
Why the test is one-sided
A one-sided superiority test focuses the hypothesis on a prespecified direction of treatment effect. Here, each posted statistical analysis identifies the hypothesis as superiority. The P-value therefore has to be interpreted in the context of that directional hypothesis rather than as a two-sided probability.
Time-to-event analysis
Overall survival and progression-free survival are time-to-event endpoints. Their analysis differs from an ordinary comparison of means because patients can have different follow-up times and some participants may not experience the event during observation. The log-rank method uses the ordering of event times and the numbers at risk over follow-up.
7. Primary Result: Overall Survival
The primary endpoint is overall survival with a 20-month time frame. The posted analysis compares Arm A (Regorafenib) with Arm B (Cetuximab, Panitumumab, Irinotecan) using a one-sided unstratified log-rank test among eligible patients.
One-sided log-rank P-value
Hypothesis: Superiority · Analysis: One-sided unstratified log-rank
Time frame: 20 months · Analysis population: Eligible patients
| Primary endpoint | Arm A | Arm B | Posted statistical result |
|---|---|---|---|
| Overall Survival | Regorafenib | Cetuximab, Panitumumab, Irinotecan | One-sided unstratified log-rank P = 0.8874 |
The reported P-value of 0.8874 is the result of the prespecified-direction superiority test reported in the registry. By itself, it does not quantify the magnitude of any difference in overall survival.
The registry does not report an OS hazard ratio, confidence interval, median OS estimate, or arm-specific OS estimate in the posted statistical analysis. Those quantities therefore cannot be used to characterize the size or precision of the treatment difference from the registry's posted analysis.
A P-value is not an effect-size measure. It describes the compatibility of the observed event-time comparison with the statistical hypothesis being tested; it does not say that one treatment changes survival by a particular percentage.
The log-rank test also does not by itself establish how large the separation between survival curves is. A hazard ratio, median survival estimate, or time-specific survival probability would provide additional information about magnitude and clinical scale when available.
8. Secondary Result: First Progression-free Survival
First progression-free survival is a secondary endpoint with an 11-month time frame. The posted analysis includes eligible patients and compares the two randomized arms using a one-sided unstratified log-rank test for superiority.
One-sided log-rank P-value
Hypothesis: Superiority · Analysis: One-sided unstratified log-rank
Time frame: 11 months · Analysis population: Eligible patients
The reported P-value of 0.9683 is the result of the one-sided superiority test. It does not measure the size of the difference in first PFS and should not be translated into a percentage treatment effect.
No PFS hazard ratio or confidence interval is reported in the posted analysis. Consequently, the precision of the relative treatment effect cannot be described from the reported statistical result alone.
For a time-to-event endpoint, the log-rank test evaluates the overall pattern of event times during follow-up. It does not replace a treatment-effect estimate. If a hazard ratio were available, its confidence interval would provide information about the range of relative effects compatible with the statistical model and data.
9. Secondary Result: Second PFS
Second PFS is a secondary endpoint with a 3-month time frame. The registry specifies a distinct analysis population: only eligible patients who experienced at least one progression were included.
One-sided log-rank P-value
Hypothesis: Superiority · Analysis: One-sided unstratified log-rank
Time frame: 3 months · Population: Eligible patients with at least one progression
The reported P-value of 0.0448 is a one-sided log-rank result for the prespecified superiority hypothesis. Its interpretation depends on the trial's prespecified type I error threshold and any multiplicity strategy; those details are not reported in the available registry information.
The analysis population is particularly important here. Because the second-PFS analysis includes only eligible patients who experienced at least one progression, it does not represent the same population as the overall-survival analysis.
No second-PFS effect estimate or confidence interval is reported. Therefore, the P-value alone does not establish how large the difference was or how precisely it was estimated.
The result should also not be treated as interchangeable with the primary OS analysis. Second PFS addresses a different point in the treatment pathway and is based on a selected subset of patients defined by progression.
10. Secondary Result: Sequential Treatment PFS
Sequential Treatment PFS is a secondary endpoint with a 12-month time frame. The posted statistical analysis includes eligible patients and uses a one-sided unstratified log-rank test for superiority.
One-sided log-rank P-value
Hypothesis: Superiority · Analysis: One-sided unstratified log-rank
Time frame: 12 months · Analysis population: Eligible patients
The reported P-value of 0.9007 is a directional hypothesis-test result. It does not quantify a treatment effect and does not indicate what percentage of patients experienced progression-free survival.
No effect estimate or confidence interval is reported for Sequential Treatment PFS. The statistical evidence available in the registry is therefore limited to the posted one-sided log-rank comparison.
As with the other time-to-event endpoints, the result should be interpreted together with the analysis population, censoring pattern, and prespecified hypothesis. A P-value alone cannot distinguish a small treatment difference from a larger difference estimated with substantial uncertainty.
11. Results Summary
| Endpoint | Role | Time frame | Method | Hypothesis | P-value |
|---|---|---|---|---|---|
| Overall Survival | Primary | 20 months | One-sided unstratified log-rank | Superiority | 0.8874 |
| First Progression-free Survival | Secondary | 11 months | One-sided unstratified log-rank | Superiority | 0.9683 |
| Second PFS | Secondary | 3 months | One-sided unstratified log-rank | Superiority | 0.0448 |
| Sequential Treatment PFS | Secondary | 12 months | One-sided unstratified log-rank | Superiority | 0.9007 |
The four posted analyses therefore report P-values rather than hazard ratios or other effect estimates. The pattern of P-values differs across endpoints, but the numerical P-values should not be interpreted as a ranking of treatments or endpoints. Each corresponds to a distinct hypothesis and analysis population or time frame.
12. Understanding the Log-Rank Test
How the comparison works
At each observed event time, the log-rank framework compares the number of events observed in each treatment group with the number expected under the null hypothesis, conditional on the patients still at risk. These contributions are then combined across the follow-up period.
The log-rank statistic aggregates these differences across event times. Censored participants remain part of the risk set until their censoring time and then no longer contribute to subsequent risk sets.
What an unstratified test means
The unstratified version treats the randomized treatment groups as the comparison units without separately weighting or conditioning the test on predefined baseline strata. The registry specifically reports the method as unstratified for OS and each secondary time-to-event analysis.
Why censoring matters
Time-to-event trials commonly contain patients who have not experienced the event when their observation ends. Such observations are censored rather than treated as events. The validity of the resulting survival comparison depends on the assumptions underlying the censoring mechanism and the way follow-up is recorded.
What the log-rank test does not provide
A log-rank P-value is not itself a measure of treatment effect. It does not report a hazard ratio, median survival, absolute survival probability, or difference in restricted mean survival time. Those quantities answer different questions and would add information about magnitude and precision.
13. Statistical Methods Explained
Why was a log-rank test used?
Overall survival and progression-free survival are time-to-event outcomes. The log-rank test is designed to compare event-time distributions while allowing participants to have different follow-up durations and allowing observations to be censored.
What does "one-sided" mean?
A one-sided test evaluates evidence in a prespecified direction. In REVERCE, the posted analyses identify the hypothesis as superiority. The reported P-value therefore corresponds to that directional hypothesis rather than to a two-sided comparison.
Why was the test unstratified?
The registry explicitly identifies the statistical method as an unstratified log-rank test. This means the reported comparison does not incorporate additional stratification factors into the log-rank calculation.
Why does the P-value not tell us the treatment effect size?
A P-value summarizes the statistical evidence against the null hypothesis under the specified test. It is influenced by the observed data and sample size, but it does not directly quantify how far apart the treatment groups are. An effect estimate and confidence interval are needed to describe magnitude and precision.
Why is second PFS analysed in a different population?
The registry specifies that only eligible patients who experienced at least one progression were included in the second-PFS analysis. This changes the population being compared and means that the result addresses a conditional stage of the treatment pathway rather than the full eligible population.
What would a hazard ratio add?
A hazard ratio would provide a relative measure of the event rate between treatment groups under a time-to-event model. A confidence interval would then describe uncertainty around that estimate. Neither quantity is reported in the posted statistical analyses for REVERCE.
14. Analysis Population and Crossover Considerations
The registry contains serious-adverse-event entries for crossover participants, identifying Crossover From Arm A: 1/4 and Crossover From Arm B: 1/8. These entries document crossover groups within the safety reporting structure.
| Group | Serious adverse events affected / at risk |
|---|---|
| Arm A: Regorafenib | 6/12 |
| Arm B: Cetuximab, Panitumumab, Irinotecan | 4/10 |
| Crossover From Arm A | 1/4 |
| Crossover From Arm B | 1/8 |
Crossover can complicate interpretation of randomized treatment comparisons because treatment received after assignment may differ from the original randomized strategy. The registry's posted statistical analyses identify the randomized-arm comparisons but do not provide a separate crossover-adjusted OS or PFS estimate.
15. Safety Results
The registry reports serious adverse events by treatment group and crossover group. The measure is presented as the number affected divided by the number at risk.
Serious adverse events
Arm A: Regorafenib · Arm B: Cetuximab, Panitumumab, Irinotecan
| Group | Affected | At risk | Reported measure |
|---|---|---|---|
| Arm A: Regorafenib | 6 | 12 | 6/12 |
| Arm B: Cetuximab, Panitumumab, Irinotecan | 4 | 10 | 4/10 |
| Crossover From Arm A | 1 | 4 | 1/4 |
| Crossover From Arm B | 1 | 8 | 1/8 |
The serious-adverse-event figures are descriptive safety data. They should not be treated as a formal hypothesis test or as an estimate of a comparative risk ratio because no such inferential analysis is reported here.
The crossover rows should also be kept separate from the primary randomized-arm safety rows. They represent groups identified by crossover status rather than additional randomized treatment arms.
16. What the P-values Do — and Do Not — Mean
The OS P-value of 0.8874 is evidence from a one-sided unstratified log-rank test of the prespecified superiority hypothesis. It does not provide an estimate of the magnitude of the OS difference.
The second-PFS P-value of 0.0448 is a one-sided log-rank result in the subset of eligible patients who experienced at least one progression. It does not establish the magnitude of the treatment effect and must be interpreted in the context of the analysis population and any prespecified multiplicity or type I error framework.
A confidence interval would provide information about precision around an effect estimate. The posted statistical analyses do not provide confidence intervals for the treatment comparisons, so precision of a relative treatment effect cannot be quantified from those results.
Effect size and statistical evidence are related but different. A P-value addresses the hypothesis test; an effect estimate addresses magnitude. A complete time-to-event interpretation normally considers both.
17. Multiplicity and Multiple Endpoints
The registry reports one primary endpoint and three secondary time-to-event endpoints. The primary endpoint is overall survival, while first PFS, second PFS, and Sequential Treatment PFS are secondary.
| Endpoint | Role | Time frame | Posted P-value |
|---|---|---|---|
| Overall Survival | Primary | 20 months | 0.8874 |
| First Progression-free Survival | Secondary | 11 months | 0.9683 |
| Second PFS | Secondary | 3 months | 0.0448 |
| Sequential Treatment PFS | Secondary | 12 months | 0.9007 |
Multiple endpoint testing raises a general statistical issue: when several hypotheses are examined, the interpretation of individual P-values depends on the prespecified type I error strategy. The available registry information does not specify a multiplicity-adjustment procedure or type I error allocation for these four analyses.
18. Non-Inferiority, Bayesian Methods, and Interim Analysis
The posted statistical analyses identify superiority as the hypothesis for overall survival, first PFS, second PFS, and Sequential Treatment PFS. A non-inferiority margin is not reported.
The registry information does not report a Bayesian statistical method, posterior probability, prior distribution, interim efficacy boundary, alpha-spending procedure, or interim-analysis result for these posted comparisons.
This distinction is important because the interpretation of a superiority trial differs from a non-inferiority trial. In a superiority analysis, the question is whether the data provide evidence in favor of a specified direction. In a non-inferiority analysis, the central question is whether the estimated effect excludes a prespecified unacceptable loss of efficacy. No non-inferiority framework is identified in the posted analyses here.
19. Censoring and Time-to-Event Interpretation
All four efficacy analyses use a time-to-event method. That means the analysis is based not only on whether an event occurred but also on when the event occurred and how long participants remained under observation.
Event information
A participant who experiences the relevant event contributes an observed event time to the analysis.
Censored information
A participant without an observed event at the end of available follow-up contributes information up to the censoring time.
Risk sets
The log-rank comparison considers participants still at risk as each event time is evaluated.
Interpretation
The resulting P-value describes a comparison of event-time distributions rather than a simple comparison of proportions.
For overall survival, the event is death. For progression-free endpoints, the exact registry endpoint definition available here is represented by the endpoint names and analysis designations, while the posted statistical analyses identify the corresponding PFS comparisons.
20. Proportional-Hazards Considerations
The posted REVERCE analyses use the log-rank test rather than reporting a Cox proportional-hazards model. This distinction matters because the log-rank test and a Cox hazard ratio answer related but different statistical questions.
A Cox model would require interpretation of a hazard ratio and would commonly rely on a proportional-hazards framework for a simple constant relative-hazard interpretation over time. No Cox model, hazard ratio, or proportional-hazards assessment is reported in the posted statistical analyses.
The absence of a reported hazard ratio also means that the magnitude of the relative treatment effect cannot be characterized from the registry's statistical-analysis results by converting the P-values into hazard ratios. A P-value and a hazard ratio are not interchangeable quantities.
21. Limitations
- Small enrollment: the trial enrolled 22 participants, so the amount of information available for time-to-event comparisons is limited.
- Open-label design: masking is listed as none. Knowledge of treatment assignment can potentially affect aspects of treatment delivery, assessment, or patient-reported outcomes, depending on the endpoint and assessment process.
- Limited effect estimates: the posted statistical analyses provide P-values but do not provide hazard ratios or confidence intervals for the treatment comparisons.
- Secondary endpoint multiplicity: three secondary time-to-event endpoints are reported, but the available registry information does not specify a multiplicity-adjustment strategy.
- Conditional second-PFS population: the second-PFS analysis includes only eligible patients who experienced at least one progression, making its population different from the primary OS population.
- Crossover: crossover groups are identified in the serious-adverse-event reporting, which can complicate interpretation of treatment-pathway outcomes.
- Unstratified analysis: the posted log-rank tests are unstratified, so no baseline-stratum adjustment is represented in the reported test.
- Missing precision measures: without confidence intervals around treatment-effect estimates, the precision and range of plausible relative effects cannot be assessed from the posted statistical analyses.
- P-value limitations: a P-value does not communicate the clinical magnitude of an effect and should not substitute for an effect estimate.
22. Why This Trial Matters Statistically
REVERCE is a useful teaching case for understanding how randomized oncology trials can use time-to-event endpoints across different stages of disease progression and treatment. The registry combines a primary overall-survival comparison with several secondary PFS analyses, each evaluated using a one-sided unstratified log-rank test.
| Concept | How it appears in REVERCE |
|---|---|
| Randomization | Participants were allocated randomly to 2 treatment arms. |
| Parallel design | The registry identifies the design model as parallel. |
| Open-label treatment | Masking is listed as none. |
| Time-to-event endpoints | OS, first PFS, second PFS, and Sequential Treatment PFS. |
| Log-rank test | All four posted statistical analyses use a one-sided unstratified log-rank method. |
| Superiority testing | Each posted analysis identifies superiority as the hypothesis. |
| Conditional analysis population | Second PFS is restricted to eligible patients who experienced at least one progression. |
| Multiplicity | One primary and three secondary time-to-event comparisons are reported. |
| Crossover | Crossover groups appear in the serious-adverse-event reporting. |
| Safety analysis | Serious adverse events are reported as affected / at risk by group. |
23. Trial Timeline
Trial start
The registry lists March 3, 2020 as the study start date.
Primary completion
The registry lists September 5, 2023 as the primary completion date.
Active, not recruiting
The study status is listed as ACTIVE_NOT_RECRUITING.
24. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary OS analysis is a one-sided unstratified log-rank superiority test with P = 0.8874. Secondary PFS analyses use the same general test framework, with reported P-values of 0.9683, 0.0448, and 0.9007.
Clinical interpretation
The registry results provide comparative hypothesis-test results and descriptive serious-adverse-event counts. They do not provide treatment-effect estimates or confidence intervals that would quantify the magnitude and precision of the efficacy differences.
The distinction is important. Statistical evidence and clinical magnitude are complementary dimensions of trial interpretation. A P-value can address a hypothesis while leaving the size of the treatment difference unresolved.
25. Reading the Primary Endpoint Correctly
The primary endpoint is overall survival at a 20-month time frame, and the registry specifies that median OS and 95% confidence intervals in each arm will be reported. The posted statistical analysis, however, identifies the one-sided unstratified log-rank comparison and provides a P-value of 0.8874.
The P-value addresses the statistical test. An effect estimate quantifies the observed treatment difference. A confidence interval describes uncertainty around an effect estimate. Keeping these quantities separate prevents a common interpretation error in clinical-trial reporting.
For example, a hazard ratio would describe relative event rates, while a median OS would describe a central feature of each survival distribution. Neither can be derived from the reported P-value without additional event-time information and assumptions.
26. Statistical Reporting Summary
| Reporting element | What the registry reports |
|---|---|
| Primary endpoint | Overall Survival (OS), 20 months |
| Primary analysis | One-sided unstratified log-rank |
| Primary hypothesis | Superiority |
| Primary P-value | 0.8874 |
| Secondary endpoints | First PFS, Second PFS, Sequential Treatment PFS |
| Secondary methods | One-sided unstratified log-rank |
| Secondary P-values | 0.9683, 0.0448, 0.9007 |
| Effect estimates | Not reported in the posted statistical analyses |
| Confidence intervals for treatment effects | Not reported in the posted statistical analyses |
| Serious adverse events | 6/12 in Arm A; 4/10 in Arm B; crossover groups also reported |
This reporting pattern illustrates why clinical-trial statistical interpretation should distinguish the existence of a hypothesis test from the availability of an interpretable effect estimate. The log-rank P-values provide formal comparisons, while the absence of hazard ratios and confidence intervals limits quantitative interpretation of treatment magnitude and precision.
27. Related Tutorials
Learn more about the methods used in this trial:
28. Related Calculators
29. Sources
- ClinicalTrials.gov: REVERCE (NCT04117945).
Continue through Clinical Biostats
Explore statistical tutorials, clinical-trial methods, and calculators for time-to-event analysis and other common biostatistical problems.
30. Record Summary
REVERCE is a randomized phase 2, open-label, parallel trial enrolling 22 participants across 2 treatment arms. Its primary endpoint is overall survival at a 20-month time frame, analyzed using a one-sided unstratified log-rank test for superiority. The posted OS P-value is 0.8874. Secondary analyses evaluate first PFS, second PFS, and Sequential Treatment PFS using the same general statistical framework, with P-values of 0.9683, 0.0448, and 0.9007, respectively.
The statistical structure is especially useful for understanding the distinction between a time-to-event hypothesis test and an effect estimate. The registry reports the log-rank P-values but does not report hazard ratios or confidence intervals for these treatment comparisons. Consequently, the reported results can be described as hypothesis-test findings without assigning an unreported magnitude or precision to the treatment effects.