This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical results on this page are restricted to the ClinicalTrials.gov record. Where the registry reports a statistical result without a confidence interval or conventional effect estimate, that limitation is preserved rather than replaced with a reconstructed value.
1. Trial at a Glance
PRIME was a completed, randomized, parallel-group phase 3 trial in metastatic colorectal cancer. The trial compared panitumumab plus FOLFOX with FOLFOX alone and reported a primary progression-free survival analysis separately in wild-type and mutant KRAS efficacy analysis sets.
| Feature | PRIME |
|---|---|
| Trial name | PRIME |
| Brief title | PRIME: Panitumumab Randomized Trial In Combination With Chemotherapy for Metastatic Colorectal Cancer to Determine Efficacy |
| Phase | Phase 3 |
| Condition | Metastatic Colorectal Cancer |
| Design | Randomized, parallel |
| Masking | None |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 1183 |
| Interventions | Panitumumab; FOLFOX regimen |
| Lead sponsor | Amgen |
| Sponsor type | Industry |
| Trial status | Completed |
| Start | 1 August 2006 |
| Primary completion | 1 August 2009 |
| ClinicalTrials.gov | NCT00364013 |
2. Clinical Question
The central question was whether adding panitumumab to the FOLFOX regimen affected progression-free survival in metastatic colorectal cancer, with the primary time-to-event analysis conducted separately according to KRAS status.
Population
Participants with metastatic colorectal cancer enrolled in the PRIME phase 3 randomized trial.
Intervention
Panitumumab plus the FOLFOX regimen.
Comparator
FOLFOX alone.
Primary question
Does panitumumab plus FOLFOX change progression-free survival relative to FOLFOX alone within the KRAS efficacy analysis sets?
3. Trial Design
Panitumumab plus FOLFOX
- Panitumumab
- FOLFOX regimen
FOLFOX alone
- FOLFOX regimen
4. Endpoints
Primary endpoint: Progression-free Survival
| Endpoint | Registry definition and time frame |
|---|---|
| Progression-free Survival | From randomization until the data cutoff date of 30 September 2008. Maximum follow-up time was 109 weeks. Progression-free survival, assessed by central radiological assessment, was defined as the time from randomization to disease progression per modified response evaluation criteria in solid tumors (RECIST) criteria or death. Participants who were alive but did not meet criteria for progression by the data cutoff date were censored at their last evaluable disease assessment date. |
Secondary endpoints with posted analyses
| Endpoint | Time frame | Analysis population | Analysis |
|---|---|---|---|
| Overall Survival | From randomization until the data cutoff date of 28 August 2009. Maximum time on follow-up was 153 weeks. | KRAS Efficacy Analysis Set | Stratified log-rank test |
| Percentage of Participants With an Objective Response | Every 8 weeks until disease progression up to the data cut-off date of 30 September 2008; Maximum follow-up time was 109 weeks. | KRAS Evaluable Central Tumor Response Analysis Set | Stratified exact test; odds ratio; adjusted for geographic region and ECOG score |
The objective-response analysis population was a subset of participants with at least one uni-dimensionally measurable lesion according to the modified RECIST criteria and blinded central radiology review.
5. Analysis Populations and KRAS Structure
The registry-reported primary PFS analyses were performed in the KRAS Efficacy Analysis Set, defined as participants for whom KRAS status was assessed. Two separate comparisons were posted: wild-type KRAS and mutant KRAS.
| Analysis set | Role | Comparison |
|---|---|---|
| KRAS Efficacy Analysis Set — wild-type | Primary PFS analysis | Wild-type KRAS: FOLFOX + panitumumab vs FOLFOX |
| KRAS Efficacy Analysis Set — mutant | Primary PFS analysis | Mutant KRAS: FOLFOX + panitumumab vs FOLFOX |
| KRAS Evaluable Central Tumor Response Analysis Set | Secondary response analysis | Wild-type or mutant KRAS treatment comparisons |
6. Statistical Methodology
Stratified log-rank testing
The primary PFS analyses and the secondary overall-survival analyses used the stratified log-rank test. The log-rank framework is designed for comparing time-to-event distributions while accounting for the fact that patients can have different follow-up times and may be censored before experiencing the event.
For a time-to-event endpoint, the test uses the ordering of event times and the numbers at risk to assess whether the treatment groups show a systematic difference in event experience over time. A stratified version performs this comparison within strata and combines the information rather than treating the entire population as a single unstratified group.
Normal score
The registry reports the effect measure for the PFS analyses as a normal score, rather than a hazard ratio. This distinction is important: the posted statistic is a hypothesis-test statistic, not a directly interpretable relative-effect measure such as an HR.
The registry explicitly states that a normal score less than zero indicates fewer than expected events for the panitumumab plus FOLFOX arm and therefore a longer progression-free survival time. A positive score indicates the opposite direction relative to the treatment arm.
Exact / Clopper-Pearson binomial methods
The objective-response analyses used a stratified exact test, normalized in the ClinicalTrials.gov record as an Exact / Clopper-Pearson binomial method. Exact methods are useful when inference for a binary endpoint should avoid relying on a large-sample normal approximation.
Odds ratios
For objective response, the registry reports an odds ratio defined as the odds of having an objective response in the panitumumab plus arm relative to the odds on the FOLFOX-alone arm.
An OR above 1 indicates higher estimated odds of response in the panitumumab-containing group; an OR below 1 indicates lower estimated odds. The OR is not a risk ratio and does not directly state how many additional participants responded.
Covariate adjustment
The objective-response analyses were adjusted for geographic region and ECOG score. Adjustment can account for prespecified or relevant covariate structure when estimating a treatment comparison, but the adjusted odds ratio should still be interpreted as a relative measure of odds rather than as an absolute response-rate difference.
7. Primary Results: Progression-Free Survival
Wild-type KRAS
Stratified log-rank normal score
P = 0.0234 · Superiority hypothesis
KRAS Efficacy Analysis Set: wild-type KRAS, FOLFOX + panitumumab vs FOLFOX
The normal score of −2.27 is in the direction specified by the registry as favorable to the panitumumab plus FOLFOX arm: a normal score below zero indicates fewer than expected events and therefore a longer progression-free survival time in that treatment group. The posted P = 0.0234 is below the stated 5% significance level for the wild-type KRAS comparison.
The statistic is not a hazard ratio. It therefore should not be described as indicating a percentage reduction in the hazard of progression or death. The registry analysis also does not report a confidence interval for this normal score, so the precision of the treatment effect cannot be expressed from the posted analysis as a conventional 95% confidence interval.
The P-value addresses compatibility with the statistical null hypothesis under the specified test; it does not measure the size of the treatment effect, clinical importance, or the probability that the treatment is beneficial. Because PFS is a censored time-to-event endpoint, interpretation also depends on the event and censoring structure underlying the stratified log-rank test.
Mutant KRAS
Stratified log-rank normal score
P = 0.0227 · Superiority hypothesis
KRAS Efficacy Analysis Set: mutant KRAS, FOLFOX + panitumumab vs FOLFOX
The mutant-KRAS normal score of 2.28 is in the opposite direction from the wild-type result. Because the registry defines a normal score below zero as indicating fewer than expected events in the panitumumab plus FOLFOX arm, a positive score indicates more events than expected in that arm relative to the FOLFOX group.
The posted P = 0.0227 is below 5%. However, the registry explicitly states that this comparison was performed conditional on first demonstrating a significant PFS treatment effect in the wild-type KRAS Efficacy Analysis Set. The wild-type comparison had P = 0.0234, satisfying the stated 5% criterion.
Again, the normal score is a test statistic rather than an effect-size measure such as a hazard ratio. No confidence interval for the normal score is reported in the posted analysis. Consequently, the result establishes the direction and statistical evidence of the comparison as reported, but does not provide a conventional interval estimate of the magnitude of the difference.
8. Secondary Results: Overall Survival
Wild-type KRAS
Stratified log-rank normal score
P = 0.0723
KRAS Efficacy Analysis Set: wild-type KRAS, FOLFOX + panitumumab vs FOLFOX
The negative normal score is directionally consistent with fewer than expected events in the panitumumab plus FOLFOX group under the registry's definition of the statistic. However, the posted P = 0.0723 is above the stated 4.99% significance level for this overall-survival comparison.
This result therefore should not be described as establishing a statistically significant overall-survival difference under the analysis criterion reported by the registry. The P-value is not a measure of effect size, and the registry-reported analysis does not provide a confidence interval or hazard ratio that would quantify the magnitude and precision of the survival difference.
The registry states that this comparison was performed conditionally on a statistically significant PFS difference in the wild-type KRAS Efficacy Analysis Set. That hierarchy is part of the interpretation and prevents the OS P-value from being treated as though it were an isolated test with no preceding condition.
Mutant KRAS
Stratified log-rank normal score
P = 0.0678
KRAS Efficacy Analysis Set: mutant KRAS, FOLFOX + panitumumab vs FOLFOX
The positive normal score is directionally opposite to the wild-type result under the registry's event-direction definition. The posted P = 0.0678 is above the stated 4.99% significance level for the mutant-KRAS overall-survival comparison.
As with the PFS analyses, the normal score should not be converted into a hazard ratio or a percentage treatment effect. The registry does not report a confidence interval for this test statistic. The comparison was also explicitly conditional on a statistically significant PFS difference in the mutant KRAS Efficacy Analysis Set.
9. Secondary Results: Objective Response
Wild-type KRAS
Odds ratio for objective response
95% CI: 0.98–1.87 · P = 0.0684
Adjusted for geographic region and ECOG score
An odds ratio of 1.35 means that the estimated odds of an objective response were 1.35 times the odds in the FOLFOX-alone group, or approximately 35% higher on an odds scale. This is not the same as saying that the response probability was 35% higher.
The 95% CI of 0.98–1.87 spans 1.00, so the interval includes the value corresponding to equal odds. The posted P = 0.0684 is also above 0.05. The confidence interval communicates precision around the odds-ratio estimate; it is not a probability that the true odds ratio lies inside the interval.
The analysis was adjusted for geographic region and ECOG score. That adjustment should be retained when interpreting the reported odds ratio rather than treating it as a crude, unadjusted comparison.
Mutant KRAS
Odds ratio for objective response
95% CI: 0.65–1.47 · P = 0.9822
Adjusted for geographic region and ECOG score
An odds ratio of 0.98 is very close to 1.00, indicating that the estimated odds of objective response were nearly the same between the panitumumab plus FOLFOX and FOLFOX-alone groups on the reported adjusted odds scale.
The 95% CI of 0.65–1.47 is compatible with both lower and higher odds of response. Its width illustrates that the estimate has meaningful uncertainty. The P = 0.9822 provides little evidence against an equal-odds null under this test, but it should not be interpreted as proving that the treatment effects are exactly identical.
As in the wild-type analysis, the odds ratio describes relative odds rather than absolute response probabilities, and it does not by itself establish whether any observed difference would be clinically important.
| Objective-response analysis | Wild-type KRAS | Mutant KRAS |
|---|---|---|
| Effect measure | Odds ratio | Odds ratio |
| Estimate | 1.35 | 0.98 |
| 95% CI | 0.98–1.87 | 0.65–1.47 |
| P-value | 0.0684 | 0.9822 |
| Adjustment | Geographic region and ECOG score | Geographic region and ECOG score |
10. Safety
The ClinicalTrials.gov record reports serious adverse events by treatment arm as the number affected divided by the number at risk.
| Safety measure | Panitumumab Plus FOLFOX | FOLFOX Alone |
|---|---|---|
| Serious adverse events, affected / at risk | 262 / 585 | 198 / 584 |
The denominators in the registry-reported serious-adverse-event result are 585 and 584. They should not be silently substituted for the overall enrollment of 1183, because the safety result is reported with its own affected/at-risk denominators.
11. How to Read the PRIME Results as a Statistical Story
PRIME is particularly instructive because the registry does not express its primary PFS findings as hazard ratios. Instead, it reports stratified log-rank normal scores and P-values. That changes what can legitimately be said about the size of the treatment effect.
Wild-type KRAS
The PFS normal score was −2.27 with P = 0.0234. Under the registry's definition, the negative score points toward fewer than expected events in the panitumumab plus FOLFOX group.
Mutant KRAS
The PFS normal score was 2.28 with P = 0.0227. The positive score points in the opposite event direction, and the test was explicitly conditional on the preceding wild-type result.
Overall survival
The wild-type and mutant normal scores were −1.80 and 1.83, with P = 0.0723 and 0.0678 respectively. Both were above the stated 4.99% significance level.
Objective response
The wild-type odds ratio was 1.35 (95% CI 0.98–1.87), whereas the mutant odds ratio was 0.98 (95% CI 0.65–1.47).
This distinction between a test statistic and an effect estimate is central. A P-value can indicate how unusual the observed test statistic would be under the null hypothesis, but it does not tell the reader how large the treatment difference is. A confidence interval around an odds ratio provides information about precision, while no such interval is posted on ClinicalTrials.gov for the reported normal scores.
12. Statistical Methods Explained
Why was a stratified log-rank test used for progression-free survival?
PFS is a time-to-event endpoint: participants can experience progression or death at different times, while some remain event-free at the data cutoff and are censored. A log-rank test compares the event-time experience between groups across follow-up rather than reducing each participant to a simple yes/no outcome. Stratification extends the comparison to account for specified strata in the analysis framework.
What does a normal score of −2.27 mean?
It is a standardized test statistic from the stratified log-rank analysis. The registry specifically states that a normal score below zero indicates fewer than expected events in the panitumumab plus FOLFOX arm and therefore a longer PFS time. It is not a hazard ratio, risk ratio, or percentage treatment effect.
Why is the mutant-KRAS analysis conditional?
The registry states that the mutant-KRAS PFS comparison was performed at a 5% significance level conditional on first demonstrating a significant treatment effect in PFS in the wild-type KRAS Efficacy Analysis Set. The wild-type analysis had P = 0.0234, so the stated condition was met before the mutant-KRAS comparison was performed.
Why is a P-value not an effect size?
A P-value summarizes evidence against a specified null hypothesis under the statistical test. It depends on both the observed departure from the null and the information available for the analysis. It therefore cannot be read as a percentage benefit, a probability that treatment works, or a measure of clinical importance. Effect size requires a measure such as an odds ratio or another directly interpretable estimate.
What does an odds ratio of 1.35 mean?
For the wild-type KRAS objective-response analysis, the OR of 1.35 means that the estimated odds of objective response with panitumumab plus FOLFOX were 1.35 times the odds with FOLFOX alone, after adjustment for geographic region and ECOG score. It does not mean that 35% more participants responded, because odds and probabilities are different quantities.
How should the 95% confidence interval be read?
The wild-type objective-response OR had a 95% CI of 0.98–1.87. The interval describes the statistical uncertainty around the estimated odds ratio under the analysis framework. Because it includes 1.00, equal odds remain compatible with the interval. The mutant-KRAS interval of 0.65–1.47 is wider and likewise includes 1.00.
13. Confidence Intervals, Censoring, and Time-to-Event Interpretation
The primary PFS analyses illustrate an important limitation of relying on the registry's posted statistical analysis alone. The registry provides the normal scores and P-values, but not conventional confidence intervals or hazard ratios for those comparisons.
What it does not provide is a conventional interval estimate of the magnitude of the treatment effect. That means the reader can interpret the direction of the test statistic and the strength of evidence under the stated hypothesis test, but cannot infer a hazard ratio or its confidence interval from the ClinicalTrials.gov record.
The endpoint itself incorporates censoring. Participants alive without qualifying progression at the data cutoff were censored at their last evaluable disease assessment date. Censoring is not equivalent to having experienced a favorable outcome: it means that the participant's event-free follow-up contributes information up to the censoring point.
For a Kaplan-Meier analysis, the survival function is conceptually built from the event experience among participants still under observation:
where di is the number of events at time ti and ni is the number at risk immediately before that time.
This equation explains why a time-to-event endpoint cannot be reconstructed from a single P-value. The underlying event times, censoring times, risk sets, and analysis strata are part of the information used by the survival analysis.
14. Multiplicity and the Conditional Testing Structure
The ClinicalTrials.gov record identifies a specific sequential structure for the primary PFS analyses. The wild-type KRAS comparison was assessed at a 5% significance level. The mutant-KRAS comparison was then assessed at 5% conditional on first demonstrating a significant wild-type PFS effect.
| Analysis | Role | Posted criterion / condition | Result |
|---|---|---|---|
| Wild-type KRAS PFS | Primary | 5% significance level | Normal score −2.27; P = 0.0234 |
| Mutant KRAS PFS | Primary | 5% significance level, conditional on significant wild-type PFS | Normal score 2.28; P = 0.0227 |
| Wild-type KRAS OS | Secondary | 4.99% significance level, conditional on significant wild-type PFS | Normal score −1.80; P = 0.0723 |
| Mutant KRAS OS | Secondary | 4.99% significance level, conditional on significant mutant PFS | Normal score 1.83; P = 0.0678 |
Conditional testing matters because the interpretation of a downstream test depends on whether the prespecified preceding condition has been met. It is therefore inappropriate to erase the sequence and present all four P-values as if they represented four unrelated tests with identical statistical roles.
15. Objective Response: Exact Testing and Covariate Adjustment
The objective-response endpoint differs fundamentally from PFS and OS because it is binary. Each evaluable participant either meets the objective-response definition or does not. The registry-reported analysis used a stratified exact test and reported odds ratios adjusted for geographic region and ECOG score.
| Feature | Wild-type KRAS | Mutant KRAS |
|---|---|---|
| Analysis population | KRAS Evaluable Central Tumor Response Analysis Set | KRAS Evaluable Central Tumor Response Analysis Set |
| Method | Stratified exact test | Stratified exact test |
| Effect measure | Odds ratio | Odds ratio |
| Adjustment | Geographic region and ECOG score | Geographic region and ECOG score |
| OR | 1.35 | 0.98 |
| 95% CI | 0.98–1.87 | 0.65–1.47 |
| P-value | 0.0684 | 0.9822 |
The exact-test framework is particularly useful for keeping the inferential calculation tied to the observed binary outcome counts rather than automatically relying on an asymptotic approximation. The covariate adjustment means that the reported odds ratios are not simply unadjusted cross-tabulations of response status.
16. Limitations
No PFS effect-size estimate
The registry supplies normal scores and P-values for PFS, but not hazard ratios or confidence intervals. The magnitude of the PFS difference therefore cannot be quantified from the registry-reported analysis as an HR.
KRAS analysis restriction
The posted primary PFS comparisons were performed in KRAS Efficacy Analysis Sets, meaning the analysis was restricted to participants for whom KRAS status was assessed.
Conditional testing
The mutant-KRAS PFS analysis and the KRAS-specific OS analyses have explicit conditional testing structures that must be retained when interpreting their P-values.
Safety comparison
Serious adverse-event counts are provided by arm, but the ClinicalTrials.gov record does not provide a formal comparative test or confidence interval for this safety endpoint.
The ClinicalTrials.gov record also do not provide baseline characteristics, subgroup forest plots, median PFS, median OS, hazard ratios, confidence intervals for the primary normal-score analyses, or a detailed missing-data or imputation strategy. Those quantities and methods are therefore not introduced here.
17. Why This Trial Matters Statistically
PRIME is a useful teaching example because it brings together several concepts that are easy to confuse when reading a clinical-trial result.
Time-to-event versus binary outcomes
PFS and OS account for event timing and censoring, whereas objective response is a binary endpoint. The two classes require different statistical summaries.
Test statistic versus effect measure
The PFS and OS analyses report normal scores rather than hazard ratios. The score is useful for hypothesis testing and direction, but it is not a directly interpretable relative treatment effect.
Hierarchical interpretation
The mutant-KRAS PFS comparison was conditional on a significant wild-type PFS result, illustrating how sequential testing can be built into a clinical-trial analysis.
Adjusted odds ratios
The response analyses demonstrate how a binary endpoint can be summarized with odds ratios while adjusting for geographic region and ECOG score.
The trial also demonstrates why a statistically literate reader should inspect the direction of the statistic, the analysis population, the testing hierarchy, and the effect measure before drawing conclusions from a P-value.
18. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
19. Related Statistical Calculators
20. Sources
- ClinicalTrials.gov: PRIME, NCT00364013.
- PubMed record: PMID 36606664.
- PubMed record: PMID 31515083.
- PubMed record: PMID 31378656.
- PubMed record: PMID 30679026.
- PubMed record: PMID 30614531.
Continue through the Clinical Biostats statistical library
Explore the statistical concepts behind randomized trials, time-to-event endpoints, binary outcomes, confidence intervals, and hypothesis testing.
21. Record Summary
PRIME provides a compact but statistically instructive example of a randomized phase 3 trial in which the principal endpoint was a censored time-to-event outcome and the registry reported stratified log-rank normal scores rather than hazard ratios. The wild-type KRAS PFS analysis produced a normal score of −2.27 with P = 0.0234, while the conditional mutant-KRAS analysis produced a score of 2.28 with P = 0.0227. Secondary overall-survival analyses produced normal scores of −1.80 and 1.83 with P-values of 0.0723 and 0.0678. Objective-response analyses used exact methods and adjusted odds ratios, with estimates of 1.35 (95% CI 0.98–1.87) in wild-type KRAS and 0.98 (95% CI 0.65–1.47) in mutant KRAS.
The central statistical lesson is that these numbers cannot all be interpreted in the same way. The normal score supplies the direction and test statistic for the stratified log-rank analysis; the P-value supplies evidence relative to the specified null; and the odds ratio supplies a relative effect measure for objective response. Reading the PRIME results correctly therefore requires attention to endpoint type, analysis population, effect measure, confidence interval, and conditional testing structure.