This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
CAIRO2 was a randomized, parallel-group phase 3 treatment trial in advanced colorectal cancer. The ClinicalTrials.gov record lists 750 participants, two treatment arms, no masking, and progression free survival and toxicity as the registered primary endpoints.
| Feature | CAIRO2 |
|---|---|
| Trial acronym | CAIRO2 |
| ClinicalTrials.gov identifier | NCT00208546 |
| Phase | Phase 3 |
| Status | COMPLETED |
| Condition | Colorectal Cancer |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 750 |
| Number of arms | 2 |
| Lead sponsor | Dutch Colorectal Cancer Group |
| Sponsor type | OTHER |
2. Clinical Question
The registry identifies CAIRO2 as a randomized phase 3 treatment trial in colorectal cancer. Its intervention field contains two regimens: capecitabine plus bevacizumab plus oxaliplatin, and capecitabine plus oxaliplatin plus bevacizumab plus cetuximab.
Population
Patients represented by the registry condition of colorectal cancer, with the study title specifying advanced colorectal cancer.
Intervention
Capecitabine + oxaliplatin + bevacizumab + cetuximab.
Comparator
Capecitabine + bevacizumab + oxaliplatin.
Primary question
How do the two randomized treatment regimens compare with respect to progression free survival and toxicity over the study duration?
The registry does not identify a superiority, non-inferiority, or equivalence framework in the information reported here. Consequently, the statistical question should be framed around the registered endpoints rather than assigning an unstated hypothesis-testing direction.
3. Trial Design
Capecitabine + oxaliplatin + bevacizumab
- Capecitabine
- Bevacizumab
- Oxaliplatin
Capecitabine + oxaliplatin + bevacizumab + cetuximab
- Capecitabine
- Oxaliplatin
- Bevacizumab
- Cetuximab
The intervention field in the registry identifies the two regimens as separate drug interventions. The record specifies randomized allocation and a parallel design, so the fundamental comparison is between treatment groups followed concurrently rather than a crossover comparison within the same participants.
4. Trial Timeline
Study start
The ClinicalTrials.gov record lists 2005-06 as the study start date.
Primary completion
The registry lists 2009-07 as the primary completion date.
Completed
The ClinicalTrials.gov record identifies the study status as COMPLETED.
5. Primary Endpoints
| Endpoint | Registry definition / time frame |
|---|---|
| Progression free survival | Progression free survival; study duration |
| Toxicity | Toxicity; study duration |
These two endpoints represent different statistical objects. Progression free survival is a time-to-event endpoint: both the time until an event and the presence of censoring matter. Toxicity is a safety outcome and is ordinarily summarized through event frequencies, proportions, severity categories, or other prespecified safety measures.
6. Planned Analysis
The ClinicalTrials.gov record lists progression free survival and toxicity as the primary endpoints but does not contain posted statistical analyses. Results have not been posted on ClinicalTrials.gov.
Progression free survival
For a randomized trial with progression free survival as an endpoint, a standard analysis begins by preserving each participant's follow-up time. Participants who experience the protocol-defined PFS event contribute an event time, while participants who reach the end of observable follow-up without the event are generally treated as right-censored.
Here, di denotes the number of events at time ti, while ni is the number of participants at risk immediately before that time.
A Kaplan-Meier curve can describe the estimated probability of remaining progression-free over time in each randomized group. A between-group time-to-event comparison can commonly be performed with a log-rank test, while a Cox proportional-hazards model can provide a hazard ratio and confidence interval when its assumptions and prespecified analysis plan support that approach.
Toxicity
Toxicity requires a different analysis structure. The primary statistical summaries would ordinarily be based on the number and proportion of participants experiencing specified toxicities, with the definition of an event, severity grading, attribution, treatment-emergent period, and analysis population determined by the protocol or statistical analysis plan.
Time-to-event endpoint
PFS requires attention to event times, censoring, follow-up duration, and the definition of progression.
Safety endpoint
Toxicity requires attention to exposure, event definitions, severity, attribution, and the population used for safety summaries.
7. Statistical Methodology
Randomization
Randomization creates the principal basis for comparing the two treatment regimens. Once participants are randomly assigned, treatment-group differences can be interpreted within the framework of a randomized comparison rather than as an observational association.
The registry identifies the allocation as RANDOMIZED and the design model as PARALLEL. This means that the statistical comparison is between independently assigned groups followed through the study rather than between sequential treatments received by the same participants.
Kaplan-Meier estimation for PFS
Progression free survival is naturally analyzed with survival-analysis methods because participants may have different lengths of follow-up and some may not experience the endpoint during observation. Kaplan-Meier estimation allows the survival function to be estimated without requiring every participant to experience the event.
A participant without an observed event can still contribute information up to the point at which follow-up ends. Treating every censored participant as though an event occurred would bias the time-to-event analysis.
Hazard-based comparison
A Cox proportional-hazards model is a common method for comparing treatment groups on a time-to-event endpoint. Its principal treatment-effect measure is the hazard ratio, which compares the modeled instantaneous event rates between groups over follow-up.
An HR below 1 indicates a lower estimated instantaneous event rate in the numerator group under the fitted model; an HR above 1 indicates a higher estimated instantaneous event rate.
A hazard ratio is not a risk ratio and is not the same as an absolute difference in progression-free survival. Its interpretation also depends on the proportional-hazards assumption when a conventional Cox model is used.
Log-rank comparison
The log-rank test provides a way to compare time-to-event distributions between randomized groups while accounting for the ordering of event times and censoring. It addresses a different question from estimation: the test assesses evidence against a specified null hypothesis, whereas the hazard ratio quantifies a relative effect under a model.
Safety analysis
Toxicity is generally evaluated separately from efficacy. Safety analyses often describe how frequently participants experience adverse events and how severe those events are. The appropriate denominator can depend on the prespecified safety population and on whether the analysis concerns all treatment-emergent events, serious events, treatment discontinuations, or a particular adverse-event category.
8. Statistical Methods Explained
Why is progression free survival a time-to-event endpoint?
Because the endpoint concerns when a participant experiences the event rather than simply whether the event ever occurs. Two participants can both experience progression while having very different progression times. Survival-analysis methods preserve that timing information.
Why use Kaplan-Meier estimation?
Not every participant necessarily has an observed PFS event during follow-up. Kaplan-Meier estimation incorporates participants who experience the event and participants whose observations are censored, allowing an estimated progression-free survival function to be constructed over time.
What does a hazard ratio describe?
A hazard ratio describes the relative instantaneous event rate between two groups under a fitted time-to-event model. It is a relative measure, not an absolute probability. For example, an HR of 0.70 would correspond to an estimated instantaneous event rate 30% lower in the numerator group, subject to the model's assumptions.
Why is the hazard ratio not the same as a risk ratio?
A risk ratio compares cumulative probabilities over a specified period. A hazard ratio compares instantaneous event rates over time. The two quantities use different statistical definitions and can differ substantially, particularly when follow-up is variable or when hazards change over time.
Why is censoring important in PFS?
Censoring determines how much follow-up information a participant contributes. A participant who remains progression-free until the end of observable follow-up should not be treated as though progression occurred on that date. The statistical method must distinguish an observed event from the end of observation.
Why should toxicity be analyzed separately from PFS?
PFS is fundamentally a time-to-event endpoint, while toxicity is generally a safety endpoint based on adverse-event occurrence and characteristics. Combining them would obscure the distinct questions being asked and could make interpretation of either endpoint less clear.
9. Interpreting the Trial's Two Primary Endpoints
The PFS endpoint asks how the randomized treatment groups differ in the time until the prespecified progression-free-survival event. Appropriate interpretation requires the event definition, censoring rules, follow-up, analysis population, and treatment-effect estimate to be considered together.
The toxicity endpoint asks about treatment-related safety over the study duration. Its interpretation depends on exactly which toxicities were counted, how severity and attribution were defined, and which participants formed the safety analysis population.
The two endpoints therefore provide complementary information. An efficacy analysis alone would not characterize the safety experience, while a toxicity summary cannot establish the time-to-event behavior of the treatment groups.
10. Statistical Interpretation of PFS
For a randomized comparison such as CAIRO2, the most informative PFS analysis would normally combine an estimated time-to-event curve with an appropriate between-group effect estimate and its uncertainty.
Absolute perspective
A Kaplan-Meier estimate at a specified time answers an absolute question: what proportion of participants are estimated to remain progression-free at that time?
Relative perspective
A hazard ratio summarizes the relative difference in modeled instantaneous event rates between the randomized groups.
Distributional perspective
The full PFS curves describe how progression-free survival evolves over the course of follow-up rather than reducing the comparison to one number.
Uncertainty perspective
Confidence intervals communicate statistical uncertainty around estimated treatment effects and should accompany point estimates when available.
These perspectives answer different questions. A hazard ratio does not provide the probability of remaining progression-free at a particular time, and a time-specific survival estimate does not by itself summarize the entire distribution of event times.
11. Statistical Interpretation of Toxicity
Toxicity requires careful definition before numerical comparisons can be interpreted. A treatment-emergent adverse event, a serious adverse event, a severe event, and an event considered related to study treatment are not interchangeable categories.
| Question | Why it matters statistically |
|---|---|
| What event is counted? | The numerator must correspond to a clearly defined safety endpoint. |
| What is the denominator? | The percentage depends on the population included in the safety analysis. |
| When is the event assessed? | Risk can depend on treatment exposure and duration of observation. |
| How is severity classified? | Combining mild and severe events can conceal clinically important differences in event intensity. |
| Is attribution required? | Treatment-related events and all adverse events represent different estimands. |
Because the registry information presented for CAIRO2 identifies toxicity as a primary endpoint but does not post a statistical analysis, no numerical toxicity comparison is reported here.
12. Multiplicity and Endpoint Hierarchy
CAIRO2 has two registered primary endpoints: progression free survival and toxicity. When a trial has more than one primary endpoint, the interpretation of statistical testing depends on how the protocol and statistical analysis plan define the relationship among those endpoints.
The registry information available here does not specify an alpha-allocation strategy, hierarchical testing procedure, multiplicity adjustment, or requirement that both endpoints meet a particular criterion for overall success.
13. Analysis Populations
The ClinicalTrials.gov information identifies randomized allocation and the two primary endpoints but does not specify the detailed efficacy or safety analysis populations in the information reported here.
For a randomized PFS analysis, an intention-to-treat population is commonly used because it preserves the treatment assignment created by randomization. Safety analyses are often based on participants who actually received treatment, because exposure is relevant to the occurrence and timing of adverse events.
The precise analysis populations, exclusions, censoring rules, and protocol deviations would ordinarily be defined prospectively in the statistical analysis plan.
14. Censoring and Missing Follow-Up
PFS analysis is especially sensitive to how incomplete follow-up is handled. A participant who has not experienced progression by the last available assessment contributes information until that observation point and may then be censored according to the prespecified rules.
Censoring means that the event has not been observed after the participant's available follow-up under the analysis rules. It does not mean that the event occurred at the censoring time.
The registry information provided for CAIRO2 does not specify detailed missing-data or imputation procedures. For a time-to-event endpoint, censoring rules are generally more central than conventional single-value imputation because the analysis is designed around observed follow-up times and event indicators.
15. Proportional-Hazards Considerations
If a Cox proportional-hazards model were used for PFS, the interpretation of a single hazard ratio would rely on the model's proportional-hazards assumption. That assumption means that the relative hazard between treatment groups is treated as approximately stable over time.
When hazards are not proportional, a single hazard ratio can compress a more complicated time-varying treatment effect into one summary measure. In such circumstances, Kaplan-Meier curves, time-specific estimates, restricted mean survival time, or other prespecified estimands may provide additional information.
16. What the Registry Reports — and Does Not Report
| Item | ClinicalTrials.gov record |
|---|---|
| Trial acronym | CAIRO2 |
| NCT identifier | NCT00208546 |
| Phase | 3 |
| Status | COMPLETED |
| Enrollment | 750 |
| Number of arms | 2 |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary endpoints | Progression free survival; toxicity |
| Primary endpoint time frame | Study duration |
| Posted statistical analyses | None |
The registry does not report numerical PFS estimates, hazard ratios, confidence intervals, p-values, toxicity event counts, or arm-specific toxicity rates in the information represented here.
17. Important Limitations and Interpretation Issues
- No posted statistical analyses: the ClinicalTrials.gov record contains no posted statistical analyses for the registered primary endpoints.
- No numerical endpoint results in the record: the registry information does not provide numerical estimates for progression free survival or toxicity.
- Endpoint definitions: the record identifies progression free survival and toxicity but does not provide the detailed operational definitions needed to reproduce an endpoint analysis.
- Analysis populations: detailed efficacy and safety population definitions are not specified in the information reported here.
- Censoring rules: the detailed PFS censoring algorithm is not reported in the available registry information.
- Multiplicity: the relationship between the two registered primary endpoints and any type I error-control strategy is not specified.
- Statistical model: the registry information does not specify whether Kaplan-Meier estimation, log-rank testing, Cox regression, or another particular PFS model was prespecified.
- Toxicity methodology: the detailed event definitions, severity criteria, attribution rules, and safety denominator are not specified in the information reported here.
18. Why This Trial Matters Statistically
CAIRO2 illustrates an important feature of clinical-trial statistics: a randomized trial can have a straightforward two-arm structure while still requiring different statistical frameworks for its primary endpoints.
| Concept | How it applies to CAIRO2 |
|---|---|
| Randomization | The registry identifies randomized allocation between two treatment arms. |
| Parallel design | The two regimens form concurrent treatment groups rather than sequential treatment periods. |
| Time-to-event analysis | Progression free survival requires analysis of event timing and censoring. |
| Kaplan-Meier estimation | A standard framework for estimating progression-free survival over time. |
| Hazard ratio | A common relative treatment-effect measure for time-to-event outcomes. |
| Log-rank testing | A common method for comparing survival distributions between randomized groups. |
| Safety analysis | Toxicity requires separate consideration of event definitions, exposure, severity, and denominators. |
| Multiplicity | Two registered primary endpoints require attention to the prespecified type I error strategy. |
| Censoring | Incomplete observation is an inherent issue in PFS analysis. |
| Analysis populations | Efficacy and safety questions can require different prespecified populations. |
19. A Statistical Reading of the CAIRO2 Design
The structure of CAIRO2 separates two important questions that are sometimes incorrectly treated as one. The first is whether the treatment regimens differ in progression-free survival. The second is how toxicity differs between them.
For PFS, the statistical object is a distribution of event times subject to censoring. This makes the analysis inherently longitudinal. A participant's contribution is not reduced simply to "progressed" or "did not progress"; the timing and amount of observed follow-up are part of the information used by survival methods.
For toxicity, the statistical object is generally a collection of safety events. The analysis must define what constitutes an event and determine an appropriate denominator. If the endpoint is summarized as a proportion, the resulting number is conditional on those definitions.
This distinction is important when interpreting randomized oncology trials. An efficacy endpoint and a safety endpoint may move in different directions, and neither one is an adequate substitute for the other.
20. Why the Absence of Posted Results Matters
The absence of posted statistical analyses means that the registry record can establish the design and planned endpoints but does not provide the numerical evidence needed to characterize the observed treatment effect.
For PFS, the missing elements would ordinarily include the observed event and censoring information or published summary estimates derived from those observations. For toxicity, numerical event counts and denominators would be required to quantify the safety experience.
What the design tells us
The trial was randomized, parallel, unmasked, phase 3, and included two treatment arms with 750 enrolled participants.
What the endpoints tell us
The registered primary outcomes were progression free survival and toxicity, both assessed over the study duration.
What a PFS analysis requires
Event times, censoring information, treatment assignment, endpoint definitions, and prespecified analysis rules.
What a toxicity analysis requires
Defined adverse-event categories, severity and attribution rules where applicable, exposure information, and appropriate analysis denominators.
21. Statistical Concepts in This Trial
The most important statistical ideas for understanding CAIRO2 are the distinction between randomized treatment comparison, time-to-event analysis, and safety analysis.
22. Sources
- ClinicalTrials.gov: CAIRO2 (NCT00208546).
Continue through the Clinical Biostats knowledge graph
Explore clinical-trial methodology, survival analysis, and statistical interpretation through the broader Clinical Biostats resource collection.
23. Record Summary
CAIRO2 is a completed phase 3 randomized parallel-group treatment trial in advanced colorectal cancer. The ClinicalTrials.gov record identifies 750 enrolled participants and two unmasked treatment arms: capecitabine plus bevacizumab plus oxaliplatin, and capecitabine plus oxaliplatin plus bevacizumab plus cetuximab.
The registered primary endpoints are progression free survival and toxicity, both with a time frame of study duration. Statistically, these endpoints require different analytical perspectives: PFS is a censored time-to-event outcome, while toxicity is a safety outcome whose interpretation depends on clearly defined events, severity, attribution, exposure, and analysis populations.
The registry does not post statistical analyses or numerical results for these primary endpoints. The appropriate statistical interpretation is therefore centered on understanding the randomized design, the distinction between efficacy and safety endpoints, and the methodology that would ordinarily be used to analyze each endpoint.