This page separates the registered trial design and endpoint from statistical interpretation. The ClinicalTrials.gov record identifies histological tumor response in the primary tumor as the primary outcome and does not report posted statistical analyses or numerical efficacy results.
1. Trial at a Glance
PRODIGE 22-ECKINOXE is a randomized phase 2 trial in locally advanced colon cancer. The registry describes three intervention strategies and identifies histological tumor response in the primary tumor, assessed using the simplified Tumor Regression Grade (TRG) of Ryan, as the primary endpoint.
| Feature | PRODIGE 22-ECKINOXE |
|---|---|
| Clinical trial | PRODIGE 22-ECKINOXE |
| NCT identifier | NCT01675999 |
| Phase | Phase 2 |
| Status | Unknown |
| Condition | Colon Cancer; Locally Advanced Malignant Neoplasm |
| Allocation | Randomized |
| Design model | Single Group |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 186 |
| Number of arms | 3 |
| Lead sponsor | Assistance Publique - Hôpitaux de Paris |
2. Clinical Question
The registry identifies a phase 2 treatment trial in locally advanced colon cancer comparing three chemotherapy strategies around surgery. The principal statistical question is whether the randomized treatment strategies differ with respect to histological tumor response in the primary tumor.
Population
Patients enrolled in a phase 2 trial for colon cancer and locally advanced malignant neoplasm.
Interventions
Perioperative simplified FOLFOX-4 chemotherapy; perioperative FOLFOX4+Cetuximab chemotherapy; or surgery followed by FOLFOX4 chemotherapy.
Comparator structure
The registry reports three intervention strategies within a randomized allocation framework rather than describing a single two-group comparison.
Primary question
How does histological tumor response in the primary tumor compare across the registered treatment strategies?
3. Trial Design
Registered intervention strategies
Perioperative simplified FOLFOX-4
- Perioperative simplified FOLFOX-4 chemotherapy
- Registry intervention type: other
Perioperative FOLFOX4 + Cetuximab
- Perioperative FOLFOX4+Cetuximab chemotherapy
- Registry intervention type: other
Surgery followed by FOLFOX4
- Surgery followed by FOLFOX4 chemotherapy
- Registry intervention type: other
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Histological tumor response in the primary tumor | Histological tumor response in the primary tumor according to the simplified Tumor Regression Grade (TRG) of Ryan | 2 years |
When the primary endpoint is evaluated
The registry specifies that the primary outcome is evaluated on surgical specimens. In the surgery-first strategy, the assessment occurs immediately after surgery. In the two strategies involving neoadjuvant chemotherapy, the assessment occurs after neoadjuvant chemotherapy (4 cycles) and surgery.
Surgery-first strategy
The registry states that the histological endpoint is evaluated immediately after surgery.
Neoadjuvant strategies
The registry states that the endpoint is evaluated after 4 cycles of neoadjuvant chemotherapy and surgery.
5. Planned Analysis
No statistical analyses are posted in the ClinicalTrials.gov record, and no numerical primary-endpoint results are reported there. The registry therefore identifies the endpoint and its assessment framework without providing an inferential result for the treatment comparison.
How a histological tumor-response endpoint is typically analysed
The simplified Tumor Regression Grade is a pathology-based response classification. Because a TRG endpoint is categorical and may retain an ordering from less regression to greater regression, the statistical analysis depends on how the categories are prespecified for the trial.
For a three-strategy randomized comparison, an analysis would ordinarily begin by displaying the number and percentage of participants in each TRG category within each treatment strategy. This preserves information about the full response distribution rather than reducing the pathology result immediately to a single binary response rate.
The basic descriptive analysis is a cross-tabulation of treatment assignment and histological response category. The appropriate inferential method depends on the prespecified definition of response and the statistical analysis plan.
If the endpoint is analysed as an ordered categorical outcome, an ordinal regression approach can be considered when its assumptions are appropriate. If the endpoint is instead dichotomized into a prespecified responder versus non-responder definition, a binary comparison or regression model may be appropriate. If the full TRG distribution is the target, collapsing categories can discard information and should be justified by the analysis plan.
What a complete analysis would need to establish
- The exact TRG categories observed in each treatment strategy.
- The number of evaluable surgical specimens in each strategy.
- The prespecified definition of histological response, if categories were collapsed into a responder endpoint.
- The statistical model or test specified for comparing the three treatment strategies.
- The analysis population and rules for participants without an evaluable surgical specimen.
- Any multiplicity procedure used for the three-strategy comparison or additional endpoints.
None of those additional numerical analysis details are reported in the ClinicalTrials.gov record summarized here. Consequently, the appropriate statistical interpretation remains centered on the registered endpoint and design rather than on an unreported treatment effect.
6. Statistical Methodology
Categorical endpoint analysis
The primary endpoint is fundamentally different from a time-to-event outcome such as overall survival. Histological tumor response is assessed from a surgical specimen and classified according to a tumor-regression grading system. The natural data structure is therefore categorical rather than continuous or survival-based.
For each treatment strategy, the distribution of participants across the TRG categories can be summarized with counts and percentages.
Three-group comparison
Because the registry reports three intervention strategies, the primary comparison is not automatically a simple two-group treatment-versus-control test. A three-group analysis can first address whether the response distributions differ overall, followed by prespecified pairwise comparisons if those comparisons are part of the statistical plan.
That distinction matters because three pairwise comparisons create more opportunities for false-positive findings than a single comparison. Any confirmatory pairwise testing should therefore follow the multiplicity strategy specified in the trial's statistical analysis plan.
Ordinal versus binary response
A tumor-regression grade can contain ordered information. Treating an ordered response as merely "response" versus "no response" may simplify interpretation but can discard distinctions among grades. Conversely, an ordinal model requires assumptions about how the ordered categories relate to treatment and about the structure of the cumulative odds.
In an ordinal logistic model, the treatment coefficient can describe a shift in the odds of being at or beyond the response categories, subject to the model's assumptions. This is a methodological example, not a statistical analysis reported by the registry.
Analysis population
For a randomized treatment comparison, the analysis population is important. An intention-to-treat analysis preserves randomized assignment and is generally central to the treatment comparison. A pathology-based endpoint introduces an additional issue: not every randomized participant necessarily has an evaluable surgical specimen. The treatment of missing or unevaluable pathology observations must therefore be prespecified rather than selected after observing the results.
Missing pathology data
If a participant does not have an evaluable surgical specimen, simply excluding that participant can change the population being compared and potentially introduce bias. Depending on the endpoint definition and estimand, the analysis plan might define such observations as non-response, use another prespecified rule, or perform a sensitivity analysis. The registry does not specify which approach was planned.
Randomization and causal interpretation
Randomized allocation provides the structural basis for comparing treatment strategies. When treatment groups are assigned randomly, systematic differences in measured and unmeasured baseline factors are expected to be balanced in expectation. The statistical analysis should preserve that randomized comparison rather than selectively redefining groups after treatment has begun.
7. Statistical Methods Explained
Why is histological tumor response an appropriate endpoint for a neoadjuvant strategy?
Neoadjuvant treatment is delivered before surgery, creating an opportunity to evaluate the resected primary tumor directly. A histological tumor-regression measure captures the pathological response observed in the surgical specimen rather than relying only on clinical or radiographic measurements.
Why does the timing of specimen assessment matter?
The registry distinguishes surgery immediately after the assigned strategy from surgery after 4 cycles of neoadjuvant chemotherapy. The specimen therefore represents the tumor state after different treatment sequences. The endpoint must be interpreted in the context of those treatment pathways.
Why might the full TRG distribution be preferable to a single response percentage?
If TRG contains multiple ordered categories, the full distribution preserves more information. Two treatment strategies could have the same percentage classified as a favorable response while differing substantially in the proportion of participants occupying intermediate categories. A complete statistical analysis should therefore begin with the category-level distribution.
How would three randomized strategies change the statistical comparison?
With three strategies, an overall comparison can test whether the response distributions differ somewhere among the groups. Specific pairwise questions are separate comparisons and can require multiplicity control if they are part of confirmatory inference.
Why is randomization important for interpreting treatment differences?
Randomization establishes the treatment groups before the outcome is observed. This helps separate differences attributable to treatment assignment from differences caused by systematic baseline selection, provided the randomized comparison is maintained in the analysis.
Why is missing pathology information a statistical issue?
The primary endpoint requires a surgical specimen. Participants without an evaluable specimen may differ from those with evaluable pathology. The handling of these observations can therefore affect the estimated treatment effect and should be defined before examining treatment-group results.
8. Design Features With Statistical Consequences
| Design feature | Registry information | Statistical consequence |
|---|---|---|
| Randomization | Randomized | Supports a comparative treatment analysis based on assigned strategy. |
| Number of arms | 3 | Requires a three-group comparison framework rather than assuming a single treatment-control contrast. |
| Masking | None | Knowledge of treatment assignment can potentially influence treatment delivery or other post-randomization processes. |
| Design model | Single group | Does not align straightforwardly with the registry's separate randomized-allocation and three-arm fields. |
| Primary purpose | Treatment | The trial is registered as a treatment study rather than a diagnostic or prevention study. |
| Primary endpoint | Histological tumor response | Requires categorical pathology analysis rather than a survival-analysis framework. |
What is not specified in the registry
The record summarized here does not report a non-inferiority margin, crossover strategy, factorial structure, Bayesian analysis, interim-analysis method, multiplicity procedure, or missing-data/imputation method. Those design features should not be inferred from the existence of randomization or from the endpoint itself.
9. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record does not post numerical statistical analyses for the primary endpoint.
- No numerical response results: the record identifies histological tumor response as the primary outcome but does not provide response counts, percentages, effect estimates, confidence intervals, or p-values.
- Design-model terminology: the registry reports randomized allocation and three arms while also reporting a single-group design model. The apparent discrepancy should be acknowledged rather than silently resolved.
- Pathology evaluability: the primary endpoint depends on surgical specimens, making the definition and handling of unevaluable specimens important to the statistical analysis.
- Three-arm multiplicity: multiple treatment comparisons can require explicit control of type I error if they are intended to support confirmatory conclusions.
- Unblinded design: the registry reports no masking. Knowledge of treatment assignment can affect aspects of trial conduct even when the primary endpoint is assessed from pathology.
- Limited registry detail: the record does not provide the complete statistical analysis plan needed to identify the exact inferential procedure for the primary endpoint.
10. Why This Trial Matters Statistically
PRODIGE 22-ECKINOXE is a useful statistical teaching case because its primary endpoint requires a different analytical framework from the time-to-event endpoints commonly seen in oncology trials. The study combines randomized treatment allocation, three intervention strategies, perioperative treatment sequencing, and a pathology-based response measure.
| Concept | How it appears in PRODIGE 22-ECKINOXE |
|---|---|
| Randomization | The registry reports randomized allocation. |
| Multi-arm design | Three intervention strategies are registered. |
| Pathology endpoint | Primary tumor response is evaluated histologically. |
| Ordinal response information | The endpoint uses the simplified Tumor Regression Grade of Ryan. |
| Timing of assessment | Surgical specimens are assessed after surgery, with different treatment sequences before surgery. |
| Missing-data considerations | The primary endpoint depends on an evaluable surgical specimen. |
| Multiplicity | Three intervention strategies create potential overall and pairwise comparisons. |
| Design consistency | Randomized allocation and three arms coexist with a registry field identifying the design model as single group. |
The deeper statistical lesson
The most important issue is not simply choosing a statistical test. It is defining the estimand first: what treatment contrast is being estimated, in which population, using which definition of histological response, and with what handling of participants whose pathology cannot be evaluated?
Once those elements are fixed, the statistical method can be selected to match the endpoint. For an ordered TRG outcome, an ordinal analysis may preserve more information than a binary response analysis. For a prespecified binary response definition, a binary model may be appropriate. For three randomized groups, the inferential structure must also distinguish an overall treatment comparison from individual pairwise contrasts.
11. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registered primary endpoint is a categorical histological response measure assessed from surgical specimens. The registry establishes the endpoint and assessment timing but does not post the statistical estimates needed to quantify differences between strategies.
Clinical interpretation
The treatment strategies differ in how chemotherapy is positioned relative to surgery, so histological tumor response is intended to characterize the pathological effect observed in the primary tumor after those treatment sequences.
12. Interpreting the Primary Endpoint Correctly
Histological tumor response describes the degree of tumor regression observed in the primary tumor specimen according to the simplified Tumor Regression Grade of Ryan. It is a pathological measure obtained from tissue after surgery.
A histological response category is not itself a direct measure of overall survival, progression-free survival, recurrence, or long-term survival. Those outcomes require their own definitions and follow-up analyses.
A treatment-group difference in TRG distribution would describe the pathological response associated with the randomized strategies. Its clinical interpretation would depend on the prespecified response definition, the magnitude and direction of the distributional difference, the precision of the estimate, and the handling of participants without evaluable specimens.
A p-value addresses compatibility with a specified null hypothesis under the statistical model. It does not describe the magnitude of the treatment difference, the probability that a treatment is effective, or the clinical importance of a pathological response difference.
13. Trial Timeline
Trial start
The ClinicalTrials.gov record lists May 2012 as the start date.
Primary completion date
The registry lists May 2012 as the primary completion date.
Unknown
The current status recorded in the ClinicalTrials.gov record is Unknown.
14. What a Complete Statistical Report Would Contain
A full statistical report for this trial would connect the randomized treatment assignment to the pathology endpoint through a prespecified analysis plan. At minimum, readers would want to see the response distribution by treatment strategy and the population used to generate those distributions.
| Component | Statistical purpose |
|---|---|
| Participants randomized to each strategy | Defines the denominators for the randomized comparison. |
| Participants with evaluable surgical specimens | Defines the population contributing directly to the pathology endpoint. |
| TRG category counts | Shows the complete observed response distribution. |
| TRG percentages | Allows response distributions to be compared across differently sized groups. |
| Prespecified response contrast | Defines the exact treatment effect being tested. |
| Confidence interval | Quantifies uncertainty around the estimated treatment difference or association. |
| Multiplicity adjustment | Controls interpretation when multiple confirmatory comparisons are made. |
| Sensitivity analysis | Examines how conclusions depend on assumptions about unevaluable or missing pathology. |
The ClinicalTrials.gov record does not provide these numerical statistical results. It does, however, identify the primary endpoint and its timing sufficiently to explain the statistical structure that an analysis would need to address.
15. Sources
- ClinicalTrials.gov: NCT01675999 — PRODIGE 22-ECKINOXE.
- PubMed record: PubMed.
- PubMed record: PubMed.
- PubMed record: PubMed.
Continue through the Clinical Biostats clinical-trial library
Clinical trial analysis connects study design, endpoint definitions, analysis populations, and statistical interpretation. Browse the broader Clinical Biostats resources for additional examples and methods.
16. Record Summary
PRODIGE 22-ECKINOXE is registered as a phase 2 treatment trial with randomized allocation, three registered intervention strategies, no masking, and an enrollment of 186 participants. Its primary endpoint is histological tumor response in the primary tumor according to the simplified Tumor Regression Grade (TRG) of Ryan, with a registry time frame of 2 years.
The endpoint is evaluated on surgical specimens: immediately after surgery for the surgery-first strategy, or after 4 cycles of neoadjuvant chemotherapy and surgery for the two neoadjuvant strategies. Statistically, this makes the trial a useful example of how categorical and potentially ordered pathology outcomes require careful endpoint definition before the inferential method is selected.
The ClinicalTrials.gov record does not post statistical analyses or numerical primary-endpoint results. The most important analytical questions are therefore structural: how the three randomized strategies are to be compared, whether the TRG outcome is analysed as an ordered categorical distribution or a prespecified binary response, how unevaluable specimens are handled, and how any multiple comparisons are controlled.