This page separates trial information reported in ClinicalTrials.gov from statistical interpretation and educational explanation. The ClinicalTrials.gov record identifies two primary overall-survival endpoints and indicate that results have been posted, but they do not provide formal statistical-analysis estimates, confidence intervals, or p-values.
1. Trial at a Glance
PARADIGM was a randomized, parallel, unmasked, phase 3 oncology trial in colorectal cancer. The registry reports 823 participants, two treatment arms, two registered primary endpoints, and results posted for the trial.
| Feature | PARADIGM |
|---|---|
| Trial name | PARADIGM |
| NCT identifier | NCT02394795 |
| Brief title | Panitumumab and RAS, Diagnostically-useful Gene Mutation for mCRC |
| Therapeutic area | Oncology |
| Condition | Colorectal Cancer |
| Phase | 3 |
| Status | COMPLETED |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 823.0 |
| Lead sponsor | Takeda |
| Sponsor type | INDUSTRY |
| Results posted | Yes |
| Outcome measures posted | 8 |
2. Clinical Question
The registered trial evaluates two treatment combinations in colorectal cancer. The central statistical question is whether overall survival differs between participants randomized to the panitumumab-containing combination and those randomized to the bevacizumab-containing combination.
Population
Participants with colorectal cancer enrolled in the phase 3 randomized trial.
Intervention
Oxaliplatin (OXA), levofolinate calcium (l-LV), 5-FU, and panitumumab.
Comparator
Oxaliplatin (OXA), levofolinate calcium (l-LV), 5-FU, and bevacizumab.
Primary question
How does overall survival compare between the two randomized treatment combinations, including among participants with left-sided tumors?
3. Trial Design
Random assignment is a core design feature of the trial.
The registry identifies a parallel-group design rather than a crossover or factorial design.
The registry identifies the trial as unmasked.
The registered primary purpose is treatment.
mFOLFOX6 + Panitumumab Combination
- Oxaliplatin (OXA)
- Levofolinate calcium (l-LV)
- 5-FU
- Panitumumab
mFOLFOX6 + Bevacizumab Combination
- Oxaliplatin (OXA)
- Levofolinate calcium (l-LV)
- 5-FU
- Bevacizumab
4. Trial Timeline
Trial start
The registered study start date was 2015-05-29.
Primary completion
The registered primary completion date was 2022-01-14.
Current registry status
The ClinicalTrials.gov record classifies the trial as COMPLETED and indicate that results have been posted.
5. Registered Primary Endpoints
The trial has two registered primary endpoints. Both concern overall survival, but one restricts the analysis population to participants with left-sided tumors while the other applies to all participants.
| Primary endpoint | Time frame | Registered definition |
|---|---|---|
| OS in Participants With Left-sided Tumors | Up to approximately 60 months | OS was measured as the time from the date of randomization to the date of death due to any cause. The left-sided tumors were defined as primary tumors occupying a left-sided site include the descending colon, sigmoid colon, and rectum. |
| Overall Survival (OS) in All Participants | Up to approximately 60 months | OS was measured as the time from the date of randomization to the date of death due to any cause. |
6. Results Availability
ClinicalTrials.gov reports that results are posted and that 8 outcome measures are posted. However, the ClinicalTrials.gov record contains no formal statistical analyses. Consequently, this page does not report a hazard ratio, confidence interval, p-value, median overall survival, survival probability, event count for the primary efficacy endpoints, or treatment-effect estimate that is not present in the ClinicalTrials.gov record.
Primary efficacy estimates registry-reported to this analysis
No formal statistical-analysis estimates, confidence intervals, or p-values are included in the ClinicalTrials.gov record.
This distinction matters. “Results posted” means that the registry contains posted outcome information, but the registry-reported statistical-analysis field does not contain the formal comparison records needed to reproduce a numerical treatment-effect summary.
7. Planned Analysis
Overall Survival in All Participants
The registered endpoint is overall survival measured from randomization to death due to any cause, with a time frame of up to approximately 60 months. For a randomized time-to-event endpoint defined this way, a standard analysis would typically estimate survival over time with the Kaplan-Meier method and compare randomized groups using a log-rank test. A Cox proportional-hazards model can be used to estimate a hazard ratio and its confidence interval when the model is appropriate.
Participants who have not experienced death by their last available assessment would ordinarily contribute follow-up information until censoring. The exact censoring rules, model specification, and estimand must come from the trial's statistical analysis plan if they are to be described as PARADIGM-specific methods.
The endpoint is naturally a time-to-event outcome because both the occurrence of death and the time from randomization to death are relevant. The ClinicalTrials.gov record does not provide the treatment-effect estimate, so no conclusion about the magnitude or direction of the observed OS difference is made here.
A future formal analysis would need to report the effect estimate together with its confidence interval. The estimate would describe the relative treatment effect under the chosen statistical model; the confidence interval would communicate precision; and a p-value, if reported, would address evidence against a specified null hypothesis rather than measure the size or clinical importance of an effect.
Overall Survival in Participants With Left-sided Tumors
The second primary endpoint uses the same overall-survival definition but restricts the population to participants whose primary tumors occupy a left-sided site, defined in the registry as the descending colon, sigmoid colon, and rectum.
Population = participants with left-sided tumors
The statistical comparison would ordinarily retain the randomized treatment groups while restricting the analysis population according to the registered left-sided-tumor definition.
A left-sided-tumor analysis is a population-restricted primary endpoint, not simply a different way of calculating the overall OS endpoint. Its interpretation depends on the prespecified definition of the left-sided population and the statistical plan governing that endpoint.
The ClinicalTrials.gov record does not report a hazard ratio, confidence interval, p-value, event count, or survival estimate for this endpoint. Therefore, no numerical treatment effect is asserted.
8. Statistical Methodology
Why Kaplan-Meier estimation is appropriate for OS
Overall survival is a time-to-event endpoint. Participants may have different follow-up times, and some participants may not have experienced the event by the end of their observed follow-up. Kaplan-Meier estimation is designed for this setting because it uses the observed event times while retaining information from participants whose observations are censored.
Here, di represents the number of events at time ti, while ni represents the number at risk immediately before that time.
Why a log-rank test is commonly used
A log-rank test provides a way to compare the survival experience of two randomized groups over follow-up. It is based on comparing observed and expected event counts over event times rather than comparing only one fixed-time survival percentage.
For PARADIGM, however, the ClinicalTrials.gov record does not state that a log-rank test was used. It is therefore presented here as the standard analytical approach for this type of registered endpoint, not as a documented trial-specific result.
Why a Cox model is commonly used
A Cox proportional-hazards model can summarize the relative event rate between treatment groups through a hazard ratio. It is particularly useful when the scientific question concerns relative differences in the rate at which the event occurs over follow-up.
An HR below 1 would indicate a lower estimated hazard in the numerator group under the fitted model; an HR above 1 would indicate a higher estimated hazard. A hazard ratio is not itself a probability, median survival, relative risk, or absolute risk difference.
Censoring is part of the analysis
Time-to-event analyses depend on how incomplete follow-up is handled. A participant who remains alive at the last observed follow-up is not equivalent to a participant who is known to have survived indefinitely. Standard survival methods therefore distinguish observed deaths from censored follow-up.
The ClinicalTrials.gov record does not specify PARADIGM's censoring rules. Those rules should not be inferred from the endpoint definition alone.
Randomization and the treatment comparison
Because the trial is registered as randomized, the primary comparison is conceptually between randomized treatment assignments rather than simply between patients who happened to receive different therapies. Randomization is intended to make the treatment groups comparable in expectation with respect to both measured and unmeasured baseline characteristics.
The ClinicalTrials.gov record does not state the randomization ratio, stratification factors, block structure, or any covariate-adjustment strategy. Those features are therefore not attributed to PARADIGM here.
9. Statistical Methods Explained
Why is overall survival a time-to-event endpoint?
Because the endpoint records both whether death occurs and when it occurs relative to randomization. Two participants can both eventually experience the event but contribute different amounts of follow-up before death. Treating OS as a simple binary endpoint would discard that timing information.
Why use randomization time as time zero?
The registry defines OS as the time from the date of randomization to the date of death due to any cause. Using randomization as time zero aligns the outcome measurement with the assignment of treatment and provides a common starting point for the randomized comparison.
What would a hazard ratio tell us?
A hazard ratio would summarize the relative instantaneous event rate between the two randomized groups under the chosen model. It would not tell us what percentage of participants benefit, how many additional months each participant lives, or the absolute difference in survival at a particular time point.
Why is a confidence interval important?
A confidence interval communicates statistical uncertainty around an estimated treatment effect. A narrow interval generally indicates greater statistical precision than a wide interval, although precision and clinical importance are separate concepts.
Why doesn't a p-value measure effect size?
A p-value addresses compatibility of the observed data with a specified null hypothesis under the statistical testing framework. It depends on both the magnitude of the observed effect and the amount of information available. Consequently, it should not be interpreted as a percentage measure of treatment benefit or as a measure of clinical importance.
Why should the left-sided analysis be interpreted separately?
The left-sided endpoint has a restricted population defined by tumor location. Its result would therefore describe the treatment comparison within that registered population. A subgroup-specific estimate cannot automatically be assumed to equal the treatment effect in all participants.
10. Left-sided Tumors and Statistical Interpretation
The PARADIGM registry defines left-sided tumors as primary tumors occupying a left-sided site including the descending colon, sigmoid colon, and rectum. This definition creates a clinically specified analysis population within the randomized trial.
Population definition
The analysis is restricted according to tumor location as defined by the registry.
Endpoint definition
OS remains time from randomization to death due to any cause.
Interpretive question
The relevant comparison is between the randomized treatment groups within the defined left-sided population.
Statistical caution
The ClinicalTrials.gov record does not report the size of the left-sided subgroup or its statistical estimate, so no numerical subgroup effect is inferred.
An important statistical distinction is between estimating an effect within a prespecified population and testing whether treatment effects differ between populations. Demonstrating effect modification generally requires an appropriate interaction analysis rather than simply observing that one subgroup has a statistically significant result while another does not.
11. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized treatment group. These are the only safety results included in the ClinicalTrials.gov record and are therefore the safety results reported here.
| Safety measure | Group P; mFOLFOX6 + Panitumumab Combinat | Group B; mFOLFOX6 + Bevacizumab Combinat |
|---|---|---|
| Serious adverse events | 115/404 | 95/407 |
12. What Is and Is Not Reported in the Supplied Data
| Topic | Supplied information | Interpretation |
|---|---|---|
| Primary OS endpoints | 2 registered endpoints | Both have results posted according to the ClinicalTrials.gov record. |
| Formal statistical analyses | None reported | No HRs, confidence intervals, or p-values can be reported from the registry-reported analysis field. |
| Median OS | Not reported | No median survival is reported here. |
| OS event counts | Not reported | No primary efficacy event count is inferred. |
| OS survival estimates | Not reported | No time-specific survival percentages are inferred. |
| Subgroup numerical results | Not reported | No subgroup effect estimate is inferred. |
| Serious adverse events | 115/404 vs 95/407 | Reported exactly as reported in the registry. |
| Outcome measures posted | 8 | The ClinicalTrials.gov record does not provide the numerical results for all 8 measures. |
13. Design Topics Not Specified in the Supplied Data
Several statistical-design topics can materially affect interpretation of a phase 3 randomized trial. The registry-reported PARADIGM data do not provide enough information to characterize them as trial-specific features.
| Design topic | Status in the ClinicalTrials.gov record | Statistical significance |
|---|---|---|
| Non-inferiority margin | Not reported | No margin should be inferred. |
| Crossover | Not reported | No crossover adjustment or dilution effect is assumed. |
| Factorial design | Not reported | The registry explicitly identifies a parallel design; no factorial structure is reported. |
| Multiplicity procedure | Not reported | No alpha-allocation or multiplicity strategy is inferred. |
| Interim analysis | Not reported | No interim-analysis schedule or stopping boundary is inferred. |
| Missing-data/imputation method | Not reported | No imputation procedure is attributed to the trial. |
| Stratification factors | Not reported | No stratified analysis model is assumed. |
| Bayesian methods | Not reported | No Bayesian analysis is attributed to the trial. |
These omissions are not minor technicalities. For example, multiplicity can affect the interpretation of p-values, stratification can affect the specified survival model, interim analyses can affect type I error control, and missing-data rules can affect estimands. When the ClinicalTrials.gov record does not specify those elements, a careful educational analysis should not fill the gaps with assumptions.
14. Primary Endpoint Interpretation Framework
The all-participants endpoint asks how long participants survive after randomization, with death from any cause defining the event. A treatment effect would ordinarily be expressed using a time-to-event measure such as a hazard ratio, supplemented by confidence intervals and, where appropriate, absolute survival estimates.
The left-sided endpoint asks the same survival question in the registry-defined population of participants with left-sided tumors. Its interpretation depends on the population definition, the randomized comparison, and the prespecified statistical analysis.
Because no formal statistical-analysis records are reported, this page cannot state whether either primary endpoint met a statistical significance threshold, what the observed hazard ratio was, how precise that estimate was, or how large the absolute survival difference was.
15. Limitations
- Incomplete numerical results: the ClinicalTrials.gov record indicates that results are posted but do not provide statistical-analysis records for the two primary endpoints.
- No effect estimates: hazard ratios, confidence intervals, p-values, median survival times, and time-specific survival estimates for OS are not reported.
- No event counts for efficacy: the number of deaths contributing to the primary OS analyses is not reported.
- Limited subgroup information: the registry definition of left-sided tumors is reported, but the corresponding subgroup size and numerical treatment effect are not.
- Safety context: serious adverse-event counts are reported, but the corresponding definitions, assessment window, and formal statistical comparison are not.
- Unspecified analysis plan: the ClinicalTrials.gov record does not identify the exact survival model, censoring rules, stratification factors, multiplicity strategy, or interim-analysis procedure.
- No treatment-effect conclusion: because the numerical statistical analyses are absent from the ClinicalTrials.gov record, this page does not infer the direction or magnitude of efficacy results.
16. Why This Trial Matters Statistically
PARADIGM provides a useful teaching case for understanding how a randomized phase 3 trial can define a time-to-event question at multiple population levels. The registry combines a broad overall-survival endpoint with a population-specific overall-survival endpoint for participants with left-sided tumors.
| Concept | How it appears in PARADIGM |
|---|---|
| Randomization | The trial is registered as RANDOMIZED. |
| Parallel design | The design model is PARALLEL. |
| Time-to-event analysis | Overall survival is defined from randomization to death due to any cause. |
| Population restriction | One primary endpoint is restricted to participants with left-sided tumors. |
| Kaplan-Meier framework | A standard method for estimating survival distributions for the registered OS endpoint. |
| Hazard ratio | A standard relative-effect measure for randomized survival comparisons, although no PARADIGM estimate is reported here. |
| Confidence interval | A standard way to communicate uncertainty around an estimated treatment effect, although no PARADIGM interval is reported here. |
| Safety analysis | Serious adverse events are reported as affected/at-risk counts by treatment group. |
| Registry-result interpretation | Results are posted, but formal statistical-analysis records are absent from the ClinicalTrials.gov record. |
The most important statistical lesson is methodological: a registry result should be interpreted using the information actually available.
17. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
18. Related Statistical Calculators
19. Sources
- ClinicalTrials.gov: PARADIGM, NCT02394795.
- Linked publication: PubMed record for PMID 38347302.
- Linked publication: PubMed record for PMID 37071094.
- Linked publication: PubMed record for PMID 28237539.
The publication links are provided because they are identified in the ClinicalTrials.gov record. Numerical trial results on this page are intentionally limited to the ClinicalTrials.gov record rather than being supplemented from those publications.
Continue with the statistical methods
Explore the survival-analysis and clinical-trial concepts that provide the statistical framework for interpreting randomized time-to-event endpoints.
20. Record Summary
PARADIGM is a completed randomized phase 3 parallel trial in colorectal cancer with 823 participants and two treatment arms: oxaliplatin, levofolinate calcium, 5-FU, and panitumumab versus oxaliplatin, levofolinate calcium, 5-FU, and bevacizumab. Its two registered primary endpoints are overall survival in participants with left-sided tumors and overall survival in all participants, both defined as time from randomization to death due to any cause and assessed for up to approximately 60 months.
The ClinicalTrials.gov record indicates that results are posted, but they do not include formal statistical-analysis records for the primary endpoints. The appropriate statistical framework for these endpoints is therefore best understood through time-to-event methodology: survival estimation, comparison of randomized groups, treatment-effect estimation, confidence intervals, and careful handling of censoring. The ClinicalTrials.gov record also report serious adverse events of 115/404 in the panitumumab combination group and 95/407 in the bevacizumab combination group.