PEAK: Panitumumab Plus mFOLFOX6 vs. Bevacizumab Plus mFOLFOX6 for First Line Treatment of Metastatic Colorectal Cancer (mCRC) Patients With Wild-Type Kirsten Rat Sarcoma-2 Virus (KRAS) Tumors: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for PEAK: Panitumumab Plus mFOLFOX6 vs. Bevacizumab Plus mFOLFOX6 for First Line Treatment of Metastatic Colorectal Cancer (mCRC) Patients With Wild-Type Kirsten Rat Sarcoma-2 Virus (KRAS) Tumors (NCT00819780), a randomized phase 2 study. The primary outcome is Progression-free Survival (PFS).

2. Clinical Question

This was a randomized comparison of two combination regimens in a defined molecular and clinical population. The question was not whether either regimen was better than no treatment: both groups received active drug combinations. The registry describes the aim as estimating the treatment effect on progression-free survival, so the result should be read as a comparison between the two assigned regimens in this trial population.

The primary estimate is compatible with a range of possible relative effects. Its confidence interval includes 1, the value corresponding to equal hazards in the model, and the reported p-value does not provide strong evidence against that reference under a conventional significance test. The estimate therefore does not establish that the regimens differ, nor does it demonstrate that they are equivalent. The interval also allows for potentially meaningful effects in either direction, so the point estimate alone should not be treated as a definitive answer.

3. Trial Design

PEAK was a randomized, multicenter, phase 2 study with a parallel-group design and no masking. The registry records 285 participants enrolled. The trial is listed as completed, with a start date of 2009-04-24 and a primary completion date of 2012-05-30.

Participants were assigned to panitumumab plus mFOLFOX6 or bevacizumab plus mFOLFOX6. The registry records 142.0 participants in the panitumumab group and 143.0 in the bevacizumab group for the primary endpoint. The primary population was intent-to-treat, meaning all randomized participants were included in the stated primary analysis set. The number analyzed can vary across other outcomes; for example, the objective-response result records 142.0 participants in each group.

The regimens shared the mFOLFOX6 chemotherapy backbone. The intervention descriptions specify oxaliplatin at 85 mg/m^2, leucovorin at 400 mg/m^2, and 5-fluorouracil at 2400 mg/m^2 on Day 1 of every 14-day cycle, alongside the assigned biologic drug. The registry descriptions are truncated at their stopping-condition language, so the complete protocol criteria for treatment discontinuation are not reported in the registry record.

The primary endpoint was progression-free survival, assessed from randomization to disease progression or a qualifying death. Investigator tumor assessments were described as occurring every 8 weeks under modified RECIST version 1.0. The primary analysis used a stratified Cox proportional hazards model. The variables used for stratification are not reported in the registry record.

The registry does not report the original sample-size justification, the planned number of events, or the assumptions used in the trial’s original design. The planning illustrations based on observed effects are separate from those original design decisions. Interim analyses and the full statistical analysis plan are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Intent-to-treat (ITT) analysis set (all randomized participants)

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as Progression-free Survival (PFS), measured over From randomization until the data cutoff date of 30 May 2012; median follow-up time was 60 weeks..

Progression-free survival (PFS) measures time from randomization until disease progression or death under the endpoint definition. In this study, progression was based on investigator assessment using modified Response Evaluation Criteria in Solid Tumors (RECIST) version 1.0. Progression included at least a 20% increase in the size of target lesions, unequivocal progression of existing non-target lesions, or any new lesions.

The registry definition also counted death within 60 days after the last evaluable tumor assessment or randomization date, whichever was later. Participants who had not met the endpoint criteria by the data cutoff were censored at the date of their last evaluable tumor assessment. The reported timeframe runs from randomization to the data cutoff date of 30 May 2012; median follow-up time was 60 weeks.

PFS is commonly used in trials of cancer treatments because it captures the time until a prespecified indicator of disease progression or death, rather than relying only on survival. That is a general explanation of the endpoint’s role; it does not mean PFS fully represents every aspect of a participant’s health or experience.

The registry reports median PFS of 10.9 months for panitumumab plus mFOLFOX6 and 10.1 months for bevacizumab plus mFOLFOX6. The comparison is reported as a hazard ratio, not as a direct test of the difference between the two medians. The hazard ratio uses event timing across follow-up, while each median describes the point at which the estimated event-free distribution reaches its midpoint.

7. Results Posted in the Registry

• Panitumumab Plus mFOLFOX6 vs Bevacizumab Plus mFOLFOX6: hazard ratio 0.87 (95% CI 0.65–1.17), reported p = 0.353 (as reported in the registry).

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Overall Survival

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Stratified Cox proportional hazards); it was not recalculated.
• Result:
  • Panitumumab Plus mFOLFOX6 vs Bevacizumab Plus mFOLFOX6: hazard ratio 0.72 (95% CI 0.47–1.11), reported p = 0.139 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval includes no difference (a ratio of 1), so these data do not show a statistically significant difference between Panitumumab Plus mFOLFOX6 vs Bevacizumab Plus mFOLFOX6 and the control. The interval gives the range of effects that remain compatible with the data.

Percentage of Participants With an Objective Response

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Bevacizumab Plus mFOLFOX6: 53.5% (76/142) versus 57.7% (82/142) on control; risk difference -4.2 percentage points (95% CI -15.8 to 7.3), risk ratio 0.93 (0.75–1.14); unadjusted two-proportion z-test p = 0.474.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Bevacizumab Plus mFOLFOX6 and the control. The interval gives the range of effects that remain compatible with the data.

Time to Disease Progression

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Stratified Cox proportional hazards); it was not recalculated.
• Result:
  • Panitumumab Plus mFOLFOX6 vs Bevacizumab Plus mFOLFOX6: hazard ratio 0.87 (95% CI 0.65–1.19), reported p = 0.386 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval includes no difference (a ratio of 1), so these data do not show a statistically significant difference between Panitumumab Plus mFOLFOX6 vs Bevacizumab Plus mFOLFOX6 and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists a broad set of secondary outcomes, including overall survival, objective response, duration of response, time to disease progression, time to initial response, resection rate, and analyses in molecular subgroups. The posted secondary results cover more than one type of outcome: time-to-event summaries and a tumor-response proportion. Across those outcomes, the estimates do not provide a single, uniformly decisive pattern that would override the primary PFS analysis.

As a general matter, secondary outcomes can help describe whether the primary result is accompanied by related patterns in survival, tumor response, or progression timing. They do not automatically confirm the primary result. Different outcomes measure different events, may have different analysis populations, and can have different precision. The available registry record does not provide result estimates for every listed secondary outcome; values not posted are not reported in the registry record.

Secondary findings also need to be interpreted in light of multiplicity. When many outcomes are examined, the chance of at least one apparently notable result can increase, even if no underlying difference exists. This is a general statistical concern. The registry record does not report whether a multiplicity adjustment was planned or applied to the secondary analyses.

Accordingly, secondary outcomes provide context but should not be treated as independent confirmatory tests unless the statistical plan supports that interpretation. Their relationship to the primary endpoint is most useful when considered with their definitions, analysis methods, confidence intervals, and the possibility of multiple testing. The primary PFS result remains the stated principal comparison in the registry record.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Overall Survival
• Percentage of Participants With an Objective Response
• Duration of Response
• Time to Disease Progression
• Time to Initial Objective Response
• Resection Rate
• Progression-free Survival (PFS) in Participants With Wild-type Rat Sarcoma Viral Oncogene Homolog (RAS)
• Progression-free Survival (PFS) in Participants With Wild-type RAS / V-raf Murine Sarcoma Viral Oncogene Homolog B1 (BRAF)

9. Statistical Interpretation

10. Statistical Methods Explained

What does the hazard ratio describe?

A hazard ratio compares the rates at which the endpoint occurs over follow-up, according to a time-to-event model. For this primary comparison, the registry reports a hazard ratio of 0.871 for panitumumab plus mFOLFOX6 relative to bevacizumab plus mFOLFOX6. A value of 1 represents equal hazards in the model; a value below 1 points toward a lower hazard in the first-named group, and a value above 1 points toward a higher hazard. This is a general explanation of the measure, not a claim that the model establishes a causal mechanism.

The hazard ratio is not a ratio of median PFS values, and it is not a direct measure of the number of months gained or lost. It uses information on event timing across participants, including participants whose follow-up ends before an event is observed. The registry names a stratified Cox proportional hazards method. The factors used to form the strata are not reported in the registry record.

What is a Cox proportional hazards analysis?

In general, a Cox proportional hazards model estimates a relative hazard while allowing the underlying hazard over time to take a flexible form. The proportional hazards part refers to the model’s assumption that the relative hazard between groups is sufficiently stable over time for a single hazard ratio to summarize the comparison. That assumption is important for interpretation. The registry record does not report diagnostics or other information that would show how the assumption was assessed.

Stratification allows the baseline hazard to differ across strata while estimating a common treatment comparison across them. That is a general description of the method. Because the registry does not identify the stratification factors, no claim can be made about which participant characteristics or design variables were used for this study.

How should the confidence interval and p-value be read?

The primary 95% confidence interval is 0.651 to 1.166. In general, a confidence interval describes the range of estimates reasonably compatible with the observed data under the method’s assumptions. It communicates uncertainty around the point estimate. It is not a range that contains a known true value with a 95% probability after the data have been observed.

The reported p-value is 0.353. In general, a p-value measures how unusual data at least as incompatible with a specified null hypothesis would be if that hypothesis and the model assumptions held. It is not a probability that the null hypothesis is true. It does not establish equivalence, and it does not say whether a possible effect would matter to patients. The registry’s result is compatible with no difference under the model, while also allowing a range of relative effects.

How do censoring and the endpoint definition matter?

In time-to-event analysis, a participant may be censored when the event has not been recorded by the end of the observation period or when the relevant outcome information ends. Censoring retains the observed follow-up time without treating the participant as having experienced the endpoint. The registry says participants who had not met the PFS criteria by the cutoff were censored at the date of their last evaluable tumor assessment.

For censoring to support valid inference, the analysis generally relies on assumptions about how censoring relates to the future event process. This is a general statistical point. Details of how the trial assessed or addressed censoring assumptions are not reported in the registry record. The endpoint also includes specified timing for deaths and uses tumor assessments at regular intervals, so its recorded event time is linked to the assessment schedule and the stated endpoint rules.

What does the intent-to-treat population contribute?

The primary analysis population is the intent-to-treat set, described as all randomized participants. In general, retaining participants according to randomized assignment helps preserve the balance created by randomization and estimates the effect of assignment to the treatment strategies. The registry identifies the population used, but it does not provide a detailed account of protocol deviations, treatment discontinuation, or how every missing outcome was handled.

Other endpoints may have different analysis populations or denominators. The posted objective-response result records 142.0 participants in each group, while the primary PFS groups record 142.0 and 143.0. Such differences should be read as outcome-specific registry reporting, not as evidence about why a participant was excluded. Reasons for any differences are not reported in the registry record.

What does the randomization contribute?

Randomization assigns participants to comparison groups by a chance process. In general, this supports a fair comparison by reducing systematic differences at assignment, although it cannot guarantee that every characteristic is balanced in a particular trial. The registry identifies the study as randomized and parallel. It does not report the randomization sequence generation, allocation concealment procedures, or the details of any stratification used in assignment.

The study was unmasked. That means the registry lists masking as none; it does not by itself establish how knowledge of assigned treatment affected outcome assessment or other study conduct. The registry describes investigator evaluation of tumor response and does not report additional procedures intended to limit assessment bias.

Why do time-to-event results differ from a simple comparison of medians?

A median is one point on an estimated event-time distribution. A hazard ratio incorporates the timing of events across the follow-up period and uses the model’s treatment comparison. The two summaries can therefore convey different aspects of the data. The registry’s primary inferential analysis is the stratified Cox model, so the hazard ratio and its confidence interval are central to the treatment comparison. The medians remain useful descriptions of the observed time-to-event distributions.

These methods are general statistical explanations. The registry record does not report the full prespecified statistical plan, model diagnostics, handling rules beyond the stated PFS censoring definition, or any planned interim analyses. Those details should not be inferred from the reported result.

11. Confidence Intervals and Precision

The primary hazard ratio is 0.871, with a 95% confidence interval from 0.651 to 1.166. The interval expresses uncertainty around the estimate: the observed data are compatible with a range of relative hazards under the analysis method and its assumptions. Because this range includes 1, the result does not rule out equal hazards in the model. It also extends below and above 1, so the direction and size of the underlying relative effect remain uncertain.

The confidence intervals for median PFS are 9.4 to 13.0 months for panitumumab plus mFOLFOX6 and 9.0 to 12.6 months for bevacizumab plus mFOLFOX6. These give uncertainty around each group’s estimated median. They are not a substitute for the confidence interval around the comparative hazard ratio. In general, separate confidence intervals for each group do not directly test the difference between groups.

Precision depends on the amount and timing of observed information, not only on the number enrolled. For time-to-event outcomes, the number and distribution of events contribute to the information available for estimating a treatment effect. The registry record does not report the event count for the primary PFS comparison in the registry record for this analysis, so no event-based explanation of the interval’s width can be made.

Confidence intervals are not guarantees about the true effect, and a wide or inclusive interval is not evidence that the groups are equivalent. The primary estimate is best read alongside its uncertainty and the stated endpoint definition. The median follow-up time was 60 weeks, which provides context for the observation period but does not by itself determine the precision of the hazard ratio.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are illustrations based on the observed effect, not the trial’s original sample-size justification. For time-to-event outcomes, the figures are total events across both groups, not participants per group. For other outcomes, the figures are participants per arm. The assumptions listed for these illustrations are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; the effect to detect is the stated fraction of the observed effect.

For the primary PFS comparison, planning at 100% of the observed effect gives 1,646 total events for 80% power and 2,204 total events for 90% power. At 75% of the observed effect, the corresponding totals are 2,926 and 3,918. At 50% of the observed effect, they are 6,584 and 8,814. These are total event counts, not enrollment targets or participants per arm.

For the objective-response comparison, the figures are participants per arm. Planning at 100% of the observed effect gives 2,170 participants per arm for 80% power and 2,904 participants per arm for 90% power. At 75% of the observed effect, the figures are 3,848 and 5,150 participants per arm; at 50%, they are 8,634 and 11,558 participants per arm.

In general, 80% power means that a study designed under the stated assumptions would have an 80% chance of detecting the specified effect at the chosen significance level if that effect were the true one. The 90% column describes the analogous chance at 90% power. Smaller effects require more participants or events to detect with the same power. These illustrations are not observed power; they are hypothetical planning calculations based on the effect used and stated assumptions. The original design assumptions and sample-size justification are not reported in the registry record.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

d is the total number of events needed (Schoenfeld's formula) with 1:1 allocation and proportional hazards.

Worked calculation for this trial

1. Hazard ratio HR = 0.871, so ln(HR) = -0.1381.
2. d = 4 × (1.96 + 0.842)² / [ln(0.871)]² = 1,645.9, rounded up to 1,646 events in total for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
hr <- 0.871                              # observed hazard ratio
events_needed <- function(power, frac = 1) ceiling(round(4 * (z_a + qnorm(power))^2 / log(hr^frac)^2, 6))   # Schoenfeld, 1:1 allocation
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: events = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(events_needed(0.80, f), big.mark = ","), format(events_needed(0.90, f), big.mark = ",")))

# Simulation: given d total events under proportional hazards, events in the treatment arm are Binomial(d, hr / (1 + hr)).
# The score (log-rank type) statistic is (D1 - d / 2) / sqrt(d / 4).
set.seed(819780)
d    <- events_needed(0.80)
reps <- 20000
D1 <- rbinom(reps, d, hr / (1 + hr))
z  <- (D1 - d / 2) / sqrt(d / 4)
cat(sprintf("Simulated power with %s events: %.3f (%d replicates)\n", format(d, big.mark = ","), mean(abs(z) > z_a), reps))

Output from running the code

100% of observed effect: events = 1,646 (80% power), 2,204 (90% power)
 75% of observed effect: events = 2,926 (80% power), 3,918 (90% power)
 50% of observed effect: events = 6,584 (80% power), 8,814 (90% power)
Simulated power with 1,646 events: 0.800 (20000 replicates)

The simulation uses the fact that, with d events in total and proportional hazards, the number of events in the treatment arm is binomial. It is an approximation that conditions on the total number of events, and it should land close to the target power.

14. Group-Sequential Design Analysis

Many time-to-event trials, especially in oncology, stop or report early when interim data are overwhelming, so a single fixed-sample number understates the real design problem. The calculation below is an illustrative group-sequential design built from this trial's posted control median and hazard ratio. The registry record does not state the planned number of looks or the spending function, so the design is a planning illustration, not a reconstruction of the protocol. Assumed: one-sided α = 0.025, 90% power, three equally spaced looks on information, Lan-DeMets O'Brien-Fleming spending, 24 months of accrual, 60 months in total and a 1% monthly dropout hazard.

Under these assumptions the design needs about 2,665 participants in total, with the boundaries below.

Fixed design versus group-sequential alternatives

Group-sequential designs cost a small increase in maximum sample size and events compared with the fixed design, in exchange for the chance to stop early for overwhelming efficacy. O'Brien-Fleming spending keeps early boundaries very conservative and the final boundary close to the fixed-design value, whereas Pocock spending spends more alpha early.

Sensitivity to a smaller true effect

Planning for exactly the observed effect is optimistic, because observed effects in published trials tend to be somewhat larger than the true effect. The table repeats the main design (three looks, O'Brien-Fleming spending, 90% power) for smaller true hazard ratios.

For context, this trial enrolled 285 participants.

Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993. Agreement between the two packages confirms the boundary calculation.

R code and output

suppressWarnings(suppressMessages({ library(gsDesign); library(rpact) }))
med_control <- 10.1          # control-arm median, months (converted from the posted unit)
hr          <- 0.871          # hazard ratio used for planning
lambdaC     <- log(2) / med_control
eta         <- 0.01               # dropout hazard per month (assumption)
acc_months  <- 24                 # accrual duration (assumption)
study_months<- 60                 # total study duration (assumption)

design <- function(k, sfu, beta, h = hr) gsSurv(k = k, test.type = 1, alpha = 0.025, beta = beta, sfu = sfu,
  lambdaC = lambdaC, hr = h, hr0 = 1, eta = eta, gamma = 1, R = acc_months, T = study_months, minfup = study_months - acc_months)

# Main design: 3 looks, Lan-DeMets O'Brien-Fleming spending, one-sided alpha 0.025, 90% power
x <- design(3, sfLDOF, 0.10)
cat(sprintf("N_TOTAL|%s\n", format(ceiling(sum(x$gamma * x$R)), big.mark = ",")))
for (i in 1:3) cat(sprintf("GS|%d|%.1f|%.0f|%.3f|%.5f|%.5f\n", i, x$T[i], x$n.I[i], x$upper$bound[i], pnorm(-x$upper$bound[i]), cumsum(x$upper$spend)[i]))
cat("\n"); print(gsBoundSummary(x))

# Comparison of designs
cmp <- list(list("Fixed design, one analysis", 1, sfLDOF), list("Two looks, O'Brien-Fleming", 2, sfLDOF),
            list("Three looks, O'Brien-Fleming", 3, sfLDOF), list("Three looks, Pocock", 3, sfLDPocock))
cat("\n")
for (b in c(0.20, 0.10)) for (d in cmp) {
  y <- design(d[[2]], d[[3]], b)
  cat(sprintf("DES|%s|%d|%.0f|%s|%.1f\n", d[[1]], round(100 * (1 - b)), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ","), tail(y$T, 1)))
}

# Sensitivity: what if the true effect is smaller than the observed one? (3 looks, O'Brien-Fleming, 90% power)
cat("\n")
for (fr in c(1, 0.75, 0.5)) {
  y <- design(3, sfLDOF, 0.10, exp(log(hr) * fr))
  cat(sprintf("SENS|%.2f|%.3f|%.0f|%s\n", fr, exp(log(hr) * fr), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ",")))
}

# Independent cross-check of the boundaries with rpact
r <- getDesignGroupSequential(kMax = 3, alpha = 0.025, beta = 0.10, sided = 1, typeOfDesign = "asOF", informationRates = x$timing)
cat(sprintf("\nCross-check, upper boundaries: gsDesign %s | rpact %s\n", paste(round(x$upper$bound, 3), collapse = ", "), paste(round(r$criticalValues, 3), collapse = ", ")))

Method: LachinFoulkes 
     Analysis               Value Efficacy
    IA 1: 33%                   Z   3.7103
      N: 1950         p (1-sided)   0.0001
  Events: 743        ~HR at bound   0.7616
    Month: 18    P(Cross) if HR=1   0.0001
              P(Cross) if HR=0.87   0.0338
    IA 2: 67%                   Z   2.5114
      N: 2666         p (1-sided)   0.0060
 Events: 1486        ~HR at bound   0.8778
    Month: 28    P(Cross) if HR=1   0.0060
              P(Cross) if HR=0.87   0.5603
        Final                   Z   1.9930
      N: 2666         p (1-sided)   0.0231
 Events: 2229        ~HR at bound   0.9190
    Month: 60    P(Cross) if HR=1   0.0250
              P(Cross) if HR=0.87   0.9000



Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993

15. Analyses Reported by the Sponsor

The registry reports the primary comparison using a stratified Cox proportional hazards analysis. The sponsor-reported hazard ratio is 0.871, its 95% confidence interval is 0.651 to 1.166, and the reported p-value is 0.353. These are the sponsor’s reported inferential results and form the main basis for describing the primary comparison.

The posted record does not flag a discrepancy between the reported analysis and the simple unadjusted recalculation. That does not turn a recalculation into a replacement for the sponsor’s method. The sponsor analysis is stratified and model-based; a simple recalculation from summary information would not necessarily reproduce that analysis or its assumptions. The exact model details beyond the named stratified Cox method are not reported in the registry record.

For time-to-event outcomes, the posted comparison uses the sponsor’s reported hazard ratio rather than deriving a new hazard ratio from the group medians. The group medians describe the distributions, while the reported Cox estimate is the treatment comparison. The same distinction applies when reading the registry’s other time-to-event comparisons.

Where a reported analysis is available, it should be read as the registry’s stated analysis, with its method, interval, and p-value. A simple recalculation can serve as a check or supplementary description, but it cannot reproduce unreported adjustments or analysis details. The registry does not provide enough information to reconstruct the full statistical analysis plan or independently verify all model choices.

16. Multiplicity, Missing Data and Analysis Populations

The registry records 14 outcome measures, 13 posted secondary outcomes, and 10 analyses. Considering multiple outcomes can increase the chance of observing a result that appears unusual even when there is no underlying treatment difference. This is the general multiplicity problem. The registry record does not report whether the trial used a prespecified hierarchy of outcomes, adjusted significance thresholds, or another multiplicity-control procedure.

The primary endpoint population is identified as the intent-to-treat analysis set, comprising all randomized participants. This is an important detail for interpreting the primary comparison. The analysis populations for all secondary endpoints are not fully specified in the registry record. The record gives outcome-specific group counts in some results, but it does not explain all differences in denominators or identify every reason for missing outcome data.

For PFS, the registry defines censoring for participants who had not met endpoint criteria by the data cutoff: they were censored at the last evaluable tumor assessment date. This states the endpoint rule, but it does not provide a complete account of missing assessments, loss to follow-up, or the assumptions used in handling those circumstances. How missing data were handled for other outcomes is not reported in the registry record.

The study is listed as unmasked, and the registry describes investigator evaluation of tumor response. Additional measures to limit assessment bias are not reported in the registry record. The record also does not report interim analyses, the complete prespecified statistical analysis plan, the sample-size justification, or model diagnostics. These omissions limit how fully the analysis process can be assessed from the registry entry alone; they do not establish that a particular procedure was or was not used outside the information recorded.

17. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: diarrhoea (6.5%), pulmonary embolism (5.0%), pyrexia (2.9%), syncope (2.9%), deep vein thrombosis (2.9%).

The registry reports serious adverse events for 139 participants at risk in each group. Serious adverse events affected 61 participants in the panitumumab plus mFOLFOX6 group and 53 in the bevacizumab plus mFOLFOX6 group. Other adverse events are listed as affecting 139 of 139 participants at risk in each group. These are reported counts, not adjusted comparisons, and the record does not provide a statistical test for differences in safety outcomes.

The registry lists diarrhoea at 6.5%, pulmonary embolism at 5.0%, pyrexia at 2.9%, syncope at 2.9%, deep vein thrombosis at 2.9%, and febrile neutropenia at 2.2% among the top serious adverse-event terms. The listed terms are not accompanied in the record by arm-specific percentages in the information, so they should not be assigned to one group or interpreted as a between-group comparison.

Death counts are not reported in the registry record. Serious adverse-event counts describe events classified as serious under the study’s reporting conventions; they do not alone capture every aspect of tolerability, severity, duration, or treatment attribution. No medical conclusion about an individual’s care follows from these aggregate figures.

18. What the Result Does, and Does Not, Mean

The primary result estimates the relative timing of progression or qualifying death for two randomized treatment combinations in participants with previously untreated, KRAS wild-type, unresectable metastatic colorectal cancer. The hazard ratio is 0.871, with a 95% confidence interval from 0.651 to 1.166 and p = 0.353. The estimate points below 1, but the interval includes 1 and permits effects on either side of that reference. The registry result therefore does not establish a difference between the regimens.

That interpretation should not be turned into a claim of equivalence. Demonstrating that two treatments are sufficiently similar requires an appropriate design and a prespecified margin for what counts as a clinically unimportant difference. Such a margin is not reported in the registry record. A result that does not demonstrate a difference can reflect uncertainty as well as similarity.

The group medians provide descriptive summaries of PFS, but they do not replace the model-based comparison. The primary result also applies to the endpoint definition, population, and follow-up recorded for the trial. It does not automatically generalize to people with other tumor characteristics, other treatment settings, or different regimens.

The secondary outcomes and safety counts add context, but their interpretation is limited by the absence of a reported multiplicity plan and by incomplete information about analysis populations and missing data. The results describe what was recorded in this trial; they do not give medical advice or settle treatment decisions for an individual. The registry record does not provide the full analysis plan or all details needed to evaluate every modeling and data-handling choice.

19. Why This Trial Matters Statistically

This trial is statistically informative because it compares two active combination regimens within a specified clinical and molecular population, rather than comparing treatment with no active regimen. Random allocation supports an interpretable comparison of assignment to those strategies. The parallel-group design and intent-to-treat primary population are relevant to understanding what effect the primary estimate targets.

The result also illustrates why an effect estimate should be read with its uncertainty. The reported hazard ratio is below 1, but the 95% confidence interval includes 1 and spans values that point in both directions. The point estimate alone would give an incomplete account of the evidence. The p-value, interval, endpoint definition, and analysis method together provide a more responsible statistical reading.

There is a distinction between the sponsor’s time-to-event analysis and simple descriptions such as group medians. The stratified Cox model uses event timing across follow-up, while medians describe a feature of each group’s event-time distribution. Understanding that distinction helps prevent treating the median values as if they were the formal comparison.

The record also shows why secondary outcomes and planning illustrations require context. Multiple secondary outcomes can raise multiplicity concerns, while planning calculations based on an observed effect are not evidence about the trial’s original design or its realized power. The registry does not report the original sample-size justification or whether multiplicity adjustments were used. These missing details limit what can be concluded about the study’s planning and confirmatory status from the registry record alone.

20. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The primary comparison is reported with a confidence interval that includes 1, so its direction and magnitude remain uncertain. The registry gives a median follow-up time of 60 weeks and a data cutoff date of 30 May 2012; the result is therefore tied to the follow-up available at that cutoff. Later follow-up or additional data are not described in the registry record.

The full statistical analysis plan, original sample-size justification, and protocol-level design assumptions are not reported in the registry record. The record names a stratified Cox proportional hazards method but does not identify the stratification factors, report diagnostics for the model assumptions, or describe interim analyses. This limits independent assessment of how the analysis was planned and implemented.

The trial was unmasked, and tumor response was evaluated by investigators. The record does not describe additional procedures for blinded assessment. This matters as context for an endpoint involving radiographic assessments, although the registry information does not establish that assessment bias occurred.

Information about missing data and analysis populations is incomplete beyond the stated intent-to-treat population for PFS and the stated PFS censoring rule. The registry does not explain all outcome-specific denominator differences or provide a comprehensive account of missing assessments. The multiplicity approach for the many outcomes is also not reported.

Finally, these results concern the two study regimens in the enrolled population. They should not be extended automatically to other populations, treatment combinations, or settings. The registry’s arm descriptions are truncated at the treatment stopping-condition language, and full protocol details are not reported in the registry record.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: PEAK: Panitumumab Plus mFOLFOX6 vs. Bevacizumab Plus mFOLFOX6 for First Line Treatment of Metastatic Colorectal Cancer (mCRC) Patients With Wild-Type Kirsten Rat Sarcoma-2 Virus (KRAS) Tumors — NCT00819780.
• Registry record last updated 2022-12-06; status last verified 2022-11.
• Results first posted 2014-08-06.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

24. Record Summary

PEAK (NCT00819780) was a completed, randomized, multicenter, phase 2, parallel-group study with no masking. It enrolled 285 participants with previously untreated, KRAS wild-type, unresectable metastatic colorectal cancer. The comparison was panitumumab plus mFOLFOX6 versus bevacizumab plus mFOLFOX6.

The primary outcome was PFS in the intent-to-treat population. The sponsor-reported stratified Cox analysis gave a hazard ratio of 0.871, a 95% confidence interval of 0.651 to 1.166, and p = 0.353. Median PFS was 10.9 months and 10.1 months, respectively. The interval includes 1, and the result does not establish a difference between the regimens.

The registry also lists secondary outcomes and serious adverse-event counts. The record does not report the original sample-size justification, multiplicity procedures, complete missing-data handling, or full statistical analysis plan.