This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
PEAK was a randomized, parallel, unmasked phase 2 oncology trial comparing panitumumab plus mFOLFOX6 with bevacizumab plus mFOLFOX6 for first-line treatment of metastatic colorectal cancer in patients with wild-type KRAS tumors. The registry reports 285 participants enrolled, two treatment arms, one registered primary endpoint, 14 posted outcome measures, and 10 posted statistical analyses.
| Feature | PEAK |
|---|---|
| Phase | Phase 2 |
| Population | Metastatic colorectal cancer patients with wild-type KRAS tumors |
| Design | Randomized, parallel, unmasked |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Primary endpoint | Progression-free Survival (PFS) |
| Primary endpoint type | Time-to-event |
| Enrollment | 285 |
| Results posted | Yes |
| Statistical analyses posted | 10 |
| Lead sponsor | Amgen |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT00819780 |
2. Clinical Question
The central question was whether first-line treatment with panitumumab plus mFOLFOX6 produced a superior progression-free survival outcome compared with bevacizumab plus mFOLFOX6 in metastatic colorectal cancer patients with wild-type KRAS tumors.
Population
Patients with colon cancer, colorectal cancer, rectal cancer, or metastatic colorectal cancer represented in the registered PEAK population, specifically described as metastatic colorectal cancer patients with wild-type KRAS tumors.
Intervention
Panitumumab plus mFOLFOX6.
Comparator
Bevacizumab plus mFOLFOX6.
Primary question
Does panitumumab plus mFOLFOX6 improve progression-free survival relative to bevacizumab plus mFOLFOX6?
3. Trial Design
Panitumumab Plus mFOLFOX6
- Panitumumab
- mFOLFOX6
Bevacizumab Plus mFOLFOX6
- Bevacizumab
- mFOLFOX6
The registry records the trial as a phase 2 randomized parallel study. The ClinicalTrials.gov record does not describe a factorial design, crossover procedure, non-inferiority margin, Bayesian analysis, or interim-analysis framework. Those topics are therefore not treated as features of the PEAK analysis on this page.
4. Trial Timeline
Trial start
The registered study start date was 24 April 2009.
Primary completion
The registered primary completion date was 30 May 2012.
Statistical analysis cutoff
The posted analyses use data from randomization until the data cutoff date of 30 May 2012, with a reported median follow-up time of 60 weeks.
5. Endpoints
| Endpoint | Registry definition / time frame | Analysis type |
|---|---|---|
| Progression-free Survival (PFS) | From randomization until the data cutoff date of 30 May 2012; median follow-up time was 60 weeks. PFS was defined as the time from the date of randomization to the date of first disease progression, or death within 60 days after the last evaluable tumor assessment or randomization date (whichever was later). Participants not meeting the criteria by the cutoff date were censored at the last evaluable tumor assessment date. | Primary time-to-event endpoint |
| Overall Survival | From randomization until the data cutoff date of 30 May 2012; median follow-up time was 60 weeks. | Secondary time-to-event endpoint |
| Time to Disease Progression | From randomization until the data cutoff date of 30 May 2012; median follow-up time was 60 weeks. | Secondary time-to-event endpoint |
| Objective Response | From randomization until the data cutoff date of 30 May 2012; median follow-up time was 60 weeks. | Secondary binary endpoint |
The registry also reports analyses restricted to participants with wild-type RAS and wild-type RAS/BRAF molecular profiles. These analyses use defined efficacy or tumor-response analysis sets rather than the full ITT population.
How PFS was defined
The registered PFS definition is more specific than the shorthand phrase "time to progression or death." Progression was based on the first disease progression, while death was incorporated when it occurred within the specified 60-day rule after the last evaluable tumor assessment or randomization date, whichever was later. Participants who had not met the event criteria by the cutoff were censored at their last evaluable tumor assessment date.
6. Statistical Methodology
Stratified Cox proportional-hazards model
The primary PFS analysis used a stratified Cox proportional-hazards model in the intent-to-treat analysis set, defined as all randomized participants. The model was stratified by prior adjuvant oxaliplatin therapy.
The hazard ratio is obtained from the fitted model as a relative comparison of the event hazard between treatment groups. Stratification allows the baseline hazard to differ across the specified stratification levels while estimating a common treatment effect across those strata.
Hazard ratio
The registry reports the Cox effect measure as a hazard ratio. For the PEAK analyses, an HR below 1.0 favors panitumumab plus mFOLFOX6 relative to bevacizumab plus mFOLFOX6.
An HR of 0.871 corresponds to an estimated hazard that is 87.1% of the comparator hazard under the fitted model. Expressed as a relative difference, 1 − 0.871 = 0.129, so the point estimate corresponds to a 12.9% lower estimated hazard. This is a derived interpretation of the reported estimate, not a separately reported trial result.
Exact / Clopper-Pearson methods for binary response
Objective-response analyses were reported using a stratified exact test, normalized in the ClinicalTrials.gov record to the exact / Clopper-Pearson binomial method category. The effect measure was an odds ratio, with stratification by prior exposure to oxaliplatin.
Intention-to-treat analysis
The primary PFS analysis used the ITT analysis set, defined in the registry data as all randomized participants. An ITT framework keeps participants associated with their randomized treatment assignment for the principal efficacy comparison.
Stratification
The primary and several secondary time-to-event analyses were stratified by prior adjuvant oxaliplatin therapy. The response analyses were also stratified, with the registry-reported analysis notes describing prior adjuvant oxaliplatin therapy or prior exposure to oxaliplatin as the stratification factor.
7. Primary Result: Progression-free Survival
The primary endpoint was progression-free survival from randomization through the 30 May 2012 data cutoff, with a reported median follow-up time of 60 weeks. The analysis was performed in the ITT population using a stratified Cox proportional-hazards model.
Primary PFS hazard ratio
95% CI: 0.651–1.166 · P = 0.3531
Two-sided 95% confidence interval; superiority hypothesis.
| Primary analysis component | Reported value |
|---|---|
| Endpoint | Progression-free Survival (PFS) |
| Analysis population | Intent-to-treat analysis set (all randomized participants) |
| Method | Stratified Cox proportional hazards |
| Stratification factor | Prior adjuvant oxaliplatin therapy |
| Effect measure | Hazard ratio |
| Estimate | 0.871 |
| 95% CI | 0.651–1.166 |
| P-value | 0.3531 |
| Hypothesis | Superiority |
What the estimate means: The reported PFS hazard ratio of 0.871 indicates that the fitted stratified Cox model estimated a lower instantaneous hazard of the PFS event for panitumumab plus mFOLFOX6 relative to bevacizumab plus mFOLFOX6. The registry specifically states that a hazard ratio below 1.0 favors panitumumab.
What it does not mean: The HR does not mean that 12.9% of patients avoided progression, nor does it mean that each individual patient had a 12.9% lower probability of progression. It is a relative model-based comparison of event hazards.
What the confidence interval says: The two-sided 95% CI of 0.651–1.166 describes uncertainty around the estimated hazard ratio under the model and sampling framework. Because the interval includes 1.0, the data are compatible with both a lower and a higher hazard under the fitted comparison.
Why the P-value is not an effect size: The P-value of 0.3531 describes the statistical evidence against the superiority null hypothesis under the specified analysis. It does not quantify how large or clinically important the treatment difference is.
Important modeling caution: The Cox model relies on its proportional-hazards framework. A single hazard ratio summarizes the relative event hazard through the model; it should not automatically be interpreted as a constant relative difference in every period of follow-up. The analysis also incorporates censoring through the time-to-event framework and uses the ITT population.
8. Secondary Results: Overall Survival
Overall survival was analyzed from randomization through the 30 May 2012 cutoff, with a median follow-up time of 60 weeks. The analysis population was the ITT analysis set and the method was a stratified Cox proportional-hazards model.
Overall survival hazard ratio
95% CI: 0.470–1.111 · P = 0.1386
Two-sided 95% confidence interval; superiority hypothesis.
| Component | Reported value |
|---|---|
| Analysis population | Intent-to-treat analysis set |
| Method | Stratified Cox proportional hazards |
| Stratification factor | Prior adjuvant oxaliplatin therapy |
| Hazard ratio | 0.723 |
| 95% CI | 0.470–1.111 |
| P-value | 0.1386 |
The point estimate is below 1.0, but the confidence interval extends above 1.0. The ClinicalTrials.gov record therefore support describing the estimate and its uncertainty without treating the result as evidence of a statistically established superiority effect under a conventional two-sided interpretation.
9. Secondary Results: Time to Disease Progression
Time to disease progression was another time-to-event outcome analyzed in the ITT population with a stratified Cox proportional-hazards model.
Time-to-progression hazard ratio
95% CI: 0.645–1.185 · P = 0.3861
Two-sided 95% confidence interval; superiority hypothesis.
| Component | Reported value |
|---|---|
| Analysis population | Intent-to-treat analysis set |
| Method | Stratified Cox proportional hazards |
| Stratification factor | Prior adjuvant oxaliplatin therapy |
| Hazard ratio | 0.874 |
| 95% CI | 0.645–1.185 |
| P-value | 0.3861 |
The PFS and time-to-progression point estimates are numerically close: 0.871 and 0.874, respectively. That similarity should not be interpreted as proof that the two endpoints are interchangeable; they have distinct registered definitions and can classify events differently.
10. Secondary Result: Objective Response
The registry reports objective response as a binary outcome measured from randomization through the 30 May 2012 data cutoff, with median follow-up of 60 weeks. The analysis used a stratified exact test, and the effect measure was the odds ratio comparing the odds of objective response in the panitumumab plus mFOLFOX6 arm with those in the bevacizumab plus mFOLFOX6 arm.
Objective-response odds ratio
95% CI: 0.72–1.95 · P = 0.5497
Two-sided 95% confidence interval; superiority hypothesis.
| Component | Reported value |
|---|---|
| Endpoint | Percentage of Participants With an Objective Response |
| Analysis population | Evaluable for Local Tumor Response Analysis Set, defined as the subset of participants in the ITT Analysis Set who had at least 1 unidimensionally measurable lesion per modified RECIST 1.0 per the registry definition |
| Method | Stratified exact test |
| Effect measure | Odds ratio |
| Estimate | 1.19 |
| 95% CI | 0.72–1.95 |
| P-value | 0.5497 |
An odds ratio of 1.19 means that the estimated odds of an objective response were 1.19 times the corresponding odds in the bevacizumab plus mFOLFOX6 group, under the specified stratified analysis.
This is not the same as saying that the response probability was 19% higher. Odds and probabilities are different quantities, and converting an odds ratio into a risk or response-rate difference requires the underlying response probabilities.
The 95% CI of 0.72–1.95 indicates substantial uncertainty around the estimated odds ratio. The P-value of 0.5497 is evidence from the specified statistical test, not a measure of the magnitude of the observed response difference.
11. Molecularly Defined Secondary Analyses
The posted registry analyses include additional time-to-event and response comparisons in participants with more narrowly defined molecular profiles. These analyses are important statistically because their analysis populations differ from the overall ITT population.
| Endpoint / population | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| PFS in participants with wild-type RAS | Stratified Cox proportional hazards | HR 0.651 | 0.444–0.956 | 0.0286 |
| PFS in participants with wild-type RAS / BRAF | Stratified Cox proportional hazards | HR 0.579 | 0.386–0.869 | 0.0083 |
| OS in participants with wild-type RAS | Stratified Cox proportional hazards | HR 0.606 | 0.337–1.088 | 0.0934 |
| OS in participants with wild-type RAS / BRAF | Stratified Cox proportional hazards | HR 0.465 | 0.239–0.902 | 0.0235 |
The ClinicalTrials.gov record defines the wild-type RAS efficacy set as a subset of the wild-type KRAS exon 2 efficacy set that includes participants with wild-type KRAS exon 2, 3, and 4 and NRAS exon 2, 3, and 4. The wild-type RAS/BRAF efficacy set is defined more narrowly to include wild-type KRAS exon 2, 3, and 4, NRAS exon 2, 3, and 4, and BRAF exon 15.
12. Objective Response in Molecularly Defined Populations
| Endpoint | Method | Odds ratio | 95% CI | P-value |
|---|---|---|---|---|
| Objective response in participants with wild-type RAS | Stratified exact test | 1.08 | 0.55–2.12 | 0.9426 |
| Objective response in participants with wild-type RAS / BRAF | Stratified exact test | 1.05 | 0.52–2.15 | 1.0000 |
The wild-type RAS objective-response analysis used a wild-type RAS investigator tumor-response analysis set, while the wild-type RAS/BRAF response analysis used a corresponding wild-type RAS/BRAF investigator tumor-response analysis set. Both were evaluated with stratified exact testing and odds ratios.
The estimates of 1.08 and 1.05 are close to 1.0, while both confidence intervals are broad and include 1.0. These results illustrate why an odds-ratio point estimate should be interpreted together with its confidence interval and the definition of the population being analyzed.
13. Results Overview
| Endpoint | Analysis population | Effect estimate | 95% CI | P-value |
|---|---|---|---|---|
| Primary PFS | ITT | HR 0.871 | 0.651–1.166 | 0.3531 |
| Overall survival | ITT | HR 0.723 | 0.470–1.111 | 0.1386 |
| Objective response | Local tumor response analysis set | OR 1.19 | 0.72–1.95 | 0.5497 |
| Time to disease progression | ITT | HR 0.874 | 0.645–1.185 | 0.3861 |
| PFS, wild-type RAS | Wild-type RAS efficacy set | HR 0.651 | 0.444–0.956 | 0.0286 |
| PFS, wild-type RAS / BRAF | Wild-type RAS/BRAF efficacy set | HR 0.579 | 0.386–0.869 | 0.0083 |
| OS, wild-type RAS | Wild-type RAS efficacy set | HR 0.606 | 0.337–1.088 | 0.0934 |
| OS, wild-type RAS / BRAF | Wild-type RAS/BRAF efficacy set | HR 0.465 | 0.239–0.902 | 0.0235 |
| Objective response, wild-type RAS | Wild-type RAS investigator tumor response set | OR 1.08 | 0.55–2.12 | 0.9426 |
| Objective response, wild-type RAS / BRAF | Wild-type RAS/BRAF investigator tumor response set | OR 1.05 | 0.52–2.15 | 1.0000 |
14. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
PFS, overall survival, and time to disease progression are time-to-event endpoints. Some participants may experience an event while others are censored because they have not experienced the event by the analysis cutoff or because follow-up ends under the endpoint rules. The Cox model is designed for this setting and provides a hazard-ratio summary of the treatment comparison without requiring a specific parametric distribution for survival time.
What does stratification by prior adjuvant oxaliplatin therapy accomplish?
The PEAK analyses stratified the Cox model by prior adjuvant oxaliplatin therapy. Stratification allows the baseline hazard function to vary across the strata while estimating the treatment effect within the stratified framework. Conceptually, the analysis therefore compares treatment groups while accounting for this prespecified factor.
Why is the primary analysis based on the ITT population?
The registry defines the primary PFS analysis population as all randomized participants. An ITT analysis preserves the treatment assignment created by randomization. This is important because randomization establishes the basis for comparing the groups without allowing post-randomization treatment decisions to redefine the principal efficacy populations.
What does an HR of 0.871 mean?
An HR of 0.871 is a relative hazard estimate. Under the fitted Cox model, the estimated event hazard for panitumumab plus mFOLFOX6 was 0.871 times the corresponding hazard for bevacizumab plus mFOLFOX6. It does not represent a 0.871-month difference, a 12.9-percentage-point difference in progression-free survival, or a 12.9% probability that an individual patient benefits.
Why does the confidence interval matter?
The point estimate is only one summary of the data. The 95% CI describes the statistical uncertainty around that estimate. For the primary PFS HR, the interval is 0.651–1.166. Because it spans 1.0, the reported estimate is compatible with both a lower and higher hazard under the model.
Why use an exact method for objective response?
Objective response is a binary outcome: participants either meet the response definition or they do not. Exact binomial procedures provide inference without relying on the same large-sample approximations used by some asymptotic methods. The registry describes the analysis as a stratified exact test and identifies the method category as Exact / Clopper-Pearson binomial.
Why is an odds ratio not the same as a response-rate ratio?
The odds of response are calculated as the probability of response divided by the probability of non-response. An odds ratio compares those odds between treatment groups. Because odds are nonlinear transformations of probabilities, an OR of 1.19 does not mean that the response percentage is 19% higher.
15. Confidence Intervals and P-values
The PEAK registry results provide two-sided 95% confidence intervals for the reported effect estimates. These intervals are especially useful for distinguishing the point estimate from the uncertainty surrounding it.
Hazard-ratio CI
For the primary PFS analysis, the HR is 0.871 with a 95% CI of 0.651–1.166. The interval includes 1.0.
Odds-ratio CI
For overall objective response, the OR is 1.19 with a 95% CI of 0.72–1.95. The interval includes 1.0.
P-value
A P-value measures evidence against a specified null hypothesis under the statistical model. It is not a measure of clinical magnitude.
Two-sided testing
The analyses posted on ClinicalTrials.gov identify their confidence intervals as two-sided and their hypothesis type as superiority.
For ratio measures such as HRs and ORs, the null value is 1.0. An interval that includes 1.0 therefore indicates that the corresponding data are not inconsistent with the null ratio under the stated confidence framework.
16. Safety Results
The ClinicalTrials.gov record provides serious adverse events by treatment arm as affected participants over participants at risk. These figures should be reported separately from the efficacy analyses because safety and efficacy address different questions and can use different analysis populations.
| Safety measure | Panitumumab Plus mFOLFOX6 | Bevacizumab Plus mFOLFOX6 |
|---|---|---|
| Serious adverse events | 61 / 139 affected / at risk | 53 / 139 affected / at risk |
The reported serious-adverse-event figures are 61/139 and 53/139. The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or P-value for serious adverse events, so no inferential safety comparison is added here.
17. Understanding the Primary PFS Result
The primary HR of 0.871 is below 1.0, which the registry specifies as favoring panitumumab. The corresponding derived relative hazard difference is 12.9% lower estimated hazard under the fitted model.
The 95% CI of 0.651–1.166 is substantially wider than a point estimate alone. It spans the null value of 1.0, so the estimate should not be presented as if the exact size of the treatment effect were known.
The two-sided P-value is 0.3531. This value does not measure treatment benefit, probability of benefit, or clinical importance. It quantifies the evidence against the specified superiority null hypothesis within the analysis framework.
The ClinicalTrials.gov record does not report median PFS, event counts, Kaplan-Meier survival probabilities at specific time points, or the underlying censoring pattern. Those quantities therefore cannot be reconstructed or added without introducing information outside the allowed trial data.
18. Secondary Molecular Analyses: How to Avoid Overinterpretation
The molecularly defined analyses show different numerical estimates across populations. For example, the wild-type RAS/BRAF PFS analysis reports an HR of 0.579, while the corresponding overall ITT PFS analysis reports an HR of 0.871.
Those estimates answer different questions because the analysis populations are different. A restricted molecular population can have a different event distribution, sample size, and amount of statistical information than the full randomized population.
19. Statistical Interpretation of the Odds Ratios
The objective-response analyses illustrate a second common clinical-trial effect measure. Unlike a hazard ratio, an odds ratio applies to a binary outcome rather than a time-to-event outcome.
Here, p denotes the probability of objective response. The registry defines its odds ratio as the odds of having an objective response in the panitumumab plus mFOLFOX6 arm relative to the odds in the bevacizumab plus mFOLFOX6 arm.
An OR of 1.19 is above 1.0, while the 95% CI of 0.72–1.95 spans both below and above 1.0. The corresponding P-value is 0.5497. Without the underlying response counts or probabilities, the odds ratio should not be converted into an absolute response-rate difference.
20. Multiplicity and Multiple Secondary Analyses
The registry reports one registered primary endpoint and nine additional posted statistical analyses, for a total of 10 statistical analyses. Several secondary analyses are molecularly restricted PFS, OS, or response comparisons.
| Analysis family | Role in the ClinicalTrials.gov record | Statistical issue |
|---|---|---|
| Progression-free survival | Registered primary endpoint | Primary superiority analysis |
| Overall survival | Secondary | Separate time-to-event comparison |
| Objective response | Secondary | Binary endpoint with exact testing |
| Time to disease progression | Secondary | Time-to-event comparison |
| Wild-type RAS / RAS-BRAF analyses | Secondary molecular analyses | Restricted populations and multiple comparisons |
The ClinicalTrials.gov record identifies the hypothesis type as superiority for the posted analyses, but they do not provide a multiplicity-adjustment strategy or alpha-allocation scheme. Accordingly, nominal P-values across the secondary analyses should not be interpreted as though the registry data establish a particular familywise-error strategy that is not provided.
21. Missing Data, Censoring, and Analysis Population
For the primary PFS endpoint, the registry explicitly describes censoring for participants who had not met the event criteria by the data cutoff: they were censored at the last evaluable tumor assessment date.
This is different from saying that their information was discarded. In a survival analysis, a censored participant contributes information about remaining event-free up to the censoring time, after which the analysis does not know the participant's subsequent event status under the specified follow-up.
ITT efficacy analysis
The primary PFS analysis uses all randomized participants, preserving randomized treatment assignment.
Response analysis set
The overall objective-response analysis uses an evaluable local tumor response population requiring at least 1 unidimensionally measurable lesion under the registry definition.
Molecular analysis sets
Wild-type RAS and wild-type RAS/BRAF analyses use progressively restricted efficacy populations.
Imputation
The ClinicalTrials.gov record does not specify a formal missing-data imputation procedure for the reported analyses.
22. Limitations
- Registry-level reporting: this analysis is limited to the ClinicalTrials.gov record. The ClinicalTrials.gov record does not include every possible statistical detail from a protocol or statistical analysis plan.
- No arm-specific enrollment count reported: although 285 participants were enrolled and there were two treatment arms, the ClinicalTrials.gov record does not state the number randomized to each arm.
- No median PFS reported: the registry reports the primary PFS hazard ratio, confidence interval, and P-value, but the ClinicalTrials.gov record does not provide a median PFS estimate.
- No event counts reported: the registry-reported primary analysis does not report the number of PFS events or the number censored.
- Censoring: PFS depends on censoring rules, including censoring at the last evaluable tumor assessment for participants not meeting event criteria by the cutoff.
- Proportional-hazards assumption: the Cox analysis summarizes treatment effect through a hazard ratio under the model's proportional-hazards framework. The ClinicalTrials.gov record does not provide a diagnostic assessment of that assumption.
- Secondary analyses: the molecularly restricted analyses address narrower populations and should not be interpreted as equivalent to the primary ITT analysis.
- Multiplicity: the ClinicalTrials.gov record does not specify a multiplicity-adjustment strategy across the 10 posted statistical analyses.
- Response population: objective response was analyzed in an evaluable tumor-response set rather than simply treating the entire ITT population as response-evaluable.
- Safety inference: serious adverse events are reported as affected/at-risk counts, but the ClinicalTrials.gov record does not provide formal statistical comparisons for these safety results.
- Missing-data methods: no formal imputation procedure is provided in the ClinicalTrials.gov record.
- Generalizability: the registered population is specifically defined around metastatic colorectal cancer and wild-type KRAS tumors, so results should not automatically be generalized beyond the represented population.
23. Why This Trial Matters Statistically
PEAK is a useful statistical teaching case because it combines a randomized treatment comparison with a primary time-to-event endpoint, stratified Cox modeling, ITT analysis, binary response analyses, exact testing, odds ratios, and molecularly restricted secondary populations.
| Concept | How it appears in PEAK |
|---|---|
| Randomization | The trial is registered as randomized with two parallel treatment arms. |
| Intention-to-treat analysis | The primary PFS analysis uses all randomized participants. |
| Time-to-event endpoint | PFS is the registered primary endpoint. |
| Hazard ratio | Used for PFS, OS, time to disease progression, and molecularly restricted survival analyses. |
| Cox proportional-hazards model | Used as a stratified model for the reported time-to-event comparisons. |
| Stratified analysis | Primary time-to-event analyses are stratified by prior adjuvant oxaliplatin therapy. |
| Censoring | PFS participants without an event by the cutoff were censored at the last evaluable tumor assessment date. |
| Confidence intervals | Two-sided 95% CIs accompany the reported HRs and ORs. |
| P-values | Reported for the primary and secondary statistical analyses. |
| Odds ratio | Used for objective-response comparisons. |
| Exact analysis | Objective response was evaluated with a stratified exact test. |
| Molecular subgroup analysis | Additional analyses were conducted in wild-type RAS and wild-type RAS/BRAF populations. |
24. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary PFS analysis produced an HR of 0.871 with a two-sided 95% CI of 0.651–1.166 and P = 0.3531. The interval includes 1.0, so the point estimate should be interpreted in the context of substantial uncertainty.
Clinical interpretation
The ClinicalTrials.gov record does not provide enough information to summarize median PFS, absolute survival probabilities, or a comprehensive clinical benefit-risk assessment. Those quantities should not be inferred from the hazard ratio alone.
The molecularly restricted results provide additional statistical information, including PFS HRs of 0.651 in the wild-type RAS population and 0.579 in the wild-type RAS/BRAF population. These are narrower analyses and should be interpreted as such rather than replacing the primary ITT result.
25. Statistical Methods Explained: A Practical Reading Guide
Because PFS measures time until a defined event, the appropriate framework accounts for both event times and censoring.
The analysis population matters because it defines which participants contribute to the principal treatment comparison.
The effect measure must match the endpoint and analysis method. A hazard ratio should not be interpreted as an odds ratio, and an odds ratio should not be interpreted as a probability difference.
No single number provides the entire statistical interpretation.
26. Related Tutorials
Learn more about the methods used in this trial:
27. Related Statistical Calculators
28. Sources
- ClinicalTrials.gov: PEAK — NCT00819780. Official registry record and source for the trial data presented on this page.
- PubMed: PMID 26049686.
Continue through Clinical Biostats
Use the related statistical tutorials and calculators to explore the survival-analysis and categorical-data methods represented in the PEAK trial.
29. Record Summary
PEAK provides a compact example of several core clinical-trial statistical methods. The trial used randomized parallel allocation and evaluated a registered primary time-to-event endpoint, progression-free survival, using an ITT analysis and a stratified Cox proportional-hazards model. The reported primary PFS HR was 0.871, with a two-sided 95% CI of 0.651–1.166 and P = 0.3531.
The same statistical framework was used for overall survival and time to disease progression, while objective response was evaluated as a binary endpoint using a stratified exact test and an odds ratio. Additional analyses examined narrower wild-type RAS and wild-type RAS/BRAF populations, illustrating the importance of distinguishing a prespecified primary population from restricted secondary analysis sets.