PRODIGE 44: Complete Statistical Analysis of Neoadjuvant mFolfirinox in Borderline Resectable Pancreatic Carcinoma

1. Trial at a Glance

PRODIGE 44 was a randomized, parallel, open-label phase 2 treatment trial in patients with borderline resectable pancreatic carcinoma. The study evaluated two neoadjuvant approaches: mFolfirinox with preoperative concomitant chemoradiotherapy and mFolfirinox without preoperative concomitant chemoradiotherapy. The registry identifies the primary efficacy endpoint as the R0 resection margin rate, assessed over a period of up to 7.5 months.

2. Clinical Question

The central question is whether adding preoperative concomitant chemoradiotherapy to neoadjuvant mFolfirinox changes the likelihood of achieving an R0 resection margin in patients with borderline resectable pancreatic carcinoma.

3. Trial Design

The intervention record lists mFolfirinox as a drug intervention, chemoradiotherapy as a radiation intervention, surgery as a procedure, and adjuvant chemotherapy as a drug intervention. The registry identifies the overall design as randomized, parallel, and unmasked.

4. Endpoints

What the R0 resection margin endpoint measures

An R0 resection is a surgical pathology classification in which the resection is considered microscopically margin-negative. The registry frames the primary endpoint as the R0 resection margin rate, so the statistical outcome is naturally expressed as a proportion of patients meeting that pathological criterion.

This makes the endpoint different from a time-to-event endpoint such as overall survival. The primary outcome is a binary surgical-pathology outcome: each relevant patient either contributes an R0 classification or does not, according to the histological assessment.

5. Planned Analysis

The registry identifies the primary endpoint but does not post a formal statistical analysis for it. No numerical estimate, confidence interval, or p-value is reported in the registry's posted statistical analyses.

For an endpoint expressed as an R0 resection margin rate, a typical analysis would compare the proportions achieving R0 resection between the randomized treatment groups. Depending on the prespecified statistical analysis plan, this could be expressed using an absolute risk difference, a relative risk, an odds ratio, or another appropriate comparison of proportions, together with a confidence interval and a hypothesis test.

6. Statistical Methodology

Randomization and between-group comparison

Because PRODIGE 44 is registered as a randomized parallel trial, the primary statistical comparison is naturally based on the difference between the randomized treatment groups. Randomization is important because it establishes the treatment assignment mechanism before the outcome is observed, supporting a causal comparison under the assumptions of the trial design.

The registry does not report a stratification scheme, covariate-adjustment strategy, or specific statistical test for the primary endpoint. Those details therefore cannot be attributed to the registered analysis.

Analysis of a binary endpoint

The R0 resection margin rate has a binary structure. For each patient in the defined analysis population, the pathological outcome can be represented as either R0 or not R0. The resulting group-specific proportions can then be compared statistically.

Confidence intervals

A confidence interval around a difference or ratio quantifies statistical uncertainty associated with the estimated treatment effect. For a binary endpoint, the interval should be interpreted in relation to the chosen effect measure: a risk-difference interval concerns an absolute difference, while a risk-ratio or odds-ratio interval concerns a relative measure.

Hypothesis testing

A conventional two-group comparison can test a null hypothesis of no difference in the selected effect measure. The resulting p-value addresses compatibility of the observed data with that null hypothesis under the specified statistical model or test. It does not measure the size or clinical importance of the observed R0 difference.

Analysis population

The denominator is especially important for a surgical endpoint. A reported R0 rate can change depending on whether the denominator includes all randomized patients or only patients who undergo resection. The registry's primary-endpoint wording identifies the R0 resection margin rate but does not provide the statistical analysis population or denominator definition.

7. Statistical Methods Explained

What is an R0 resection margin rate?

The R0 resection margin rate is the proportion of patients whose surgical specimen is classified as having a microscopically negative resection margin. It is therefore naturally analyzed as a binary or proportion outcome rather than as a continuous measurement.

Why is randomization important for an R0 endpoint?

Randomization determines treatment assignment independently of subsequent outcome information, subject to adherence to the study protocol and other trial assumptions. This makes a between-group comparison of the R0 rate a comparison of randomized treatment strategies rather than merely a comparison of patients who happened to receive different therapies.

What does an absolute risk difference tell us?

An absolute risk difference describes the percentage-point separation between two R0 rates. For example, if one group had an R0 rate of 60% and another had 50%, the risk difference would be 10 percentage points. That example illustrates the interpretation only; it is not a result from PRODIGE 44.

What does a risk ratio tell us?

A risk ratio compares the probability of R0 resection in one treatment group with the probability in the other. A value above 1 would indicate a higher estimated R0 probability in the numerator group, while a value below 1 would indicate a lower estimated probability.

Why does the denominator matter?

Suppose the R0 endpoint is calculated only among patients who undergo surgery. The resulting rate answers a different question from an analysis in which all randomized patients remain in the denominator and patients who do not undergo resection are assigned a prespecified non-R0 outcome. For a randomized trial, the exact analysis population should therefore be defined before examining the results.

Why is a confidence interval useful?

A confidence interval provides information about the precision of the estimated treatment effect. A narrow interval indicates greater statistical precision than a wide interval, while the location of the interval relative to the null value depends on whether the analysis uses a difference, ratio, or another effect measure.

Why should the R0 endpoint not be treated as a survival endpoint?

R0 resection status is determined from surgical pathology, whereas survival endpoints describe time until an event. R0 rate therefore does not require Kaplan-Meier estimation or a Cox proportional-hazards model simply because the trial is a clinical oncology study. The statistical method should follow the structure of the endpoint.

8. Statistical Interpretation of the Primary Endpoint

9. Endpoint-Specific Interpretation Issues

10. Trial Timeline

11. What the Registry Reports — and What It Does Not

The absence of a posted statistical analysis means that the registry record establishes the study design and primary endpoint but does not provide a numerical statistical result for the R0 resection margin rate.

12. Limitations

• No posted statistical analysis: the ClinicalTrials.gov record does not contain a formal posted statistical analysis for the primary endpoint, so no treatment-effect estimate, confidence interval, or p-value can be reported from the registry.
• Analysis population not specified in the available record: the registry identifies the R0 resection margin rate but does not specify the denominator or detailed analysis population for that endpoint.
• No reported stratification details: the available record does not identify randomization strata or a covariate-adjusted analysis strategy.
• No reported hypothesis-testing framework: the registry does not specify the statistical test, effect measure, confidence-interval method, or multiplicity procedure for the primary endpoint.
• Binary endpoint interpretation: R0 resection status captures a surgical-pathology outcome and should not be interpreted as a direct surrogate for survival without additional evidence.
• Two-treatment comparison: the trial evaluates two neoadjuvant strategies, but the available record does not provide numerical arm-specific R0 rates.
• Limited endpoint scope: the registered primary endpoint is specifically the histological R0 resection margin rate, so conclusions based on this record should remain focused on that outcome rather than extending to unreported clinical outcomes.

13. Why This Trial Matters Statistically

PRODIGE 44 is a useful statistical teaching case because it illustrates an important distinction in clinical-trial methodology: The statistical method should be selected from the structure of the endpoint, not simply from the disease area or the fact that the study is randomized.

14. Statistical Interpretation vs Clinical Interpretation

15. Why the Analysis Population Deserves Attention

For a surgical endpoint, the denominator can materially change the meaning of the reported rate. Consider two conceptual approaches. One approach could retain all randomized patients in the denominator and apply a prespecified rule to patients who do not reach surgery. Another could restrict the denominator to patients who undergo resection and have an assessable pathological specimen.

Those approaches answer different questions. The first is closer to an overall randomized-strategy perspective, while the second describes the pathological outcome among patients who reach the relevant surgical assessment. Neither approach should be assumed to represent the PRODIGE 44 analysis without an explicit statistical-analysis definition.

16. Statistical Interpretation of Possible Effect Measures

These measures are mathematically related but are not interchangeable. When the event probability is not rare, an odds ratio can differ substantially in numerical interpretation from a risk ratio. For that reason, the selected effect measure should always be stated explicitly.

17. Understanding Confidence Intervals for a Proportion Comparison

For PRODIGE 44, a confidence interval around an R0 treatment effect would be more informative than a point estimate alone. If the primary comparison were expressed as a risk difference, the interval would be interpreted on an absolute percentage-point scale. If expressed as a ratio, the null value would be 1 rather than 0.

The confidence interval should not be interpreted as a range containing the true treatment effect with a fixed probability after the data have been observed. It is a statement about the performance of the specified confidence procedure under repeated sampling.

18. Why a P-Value Would Not Be Enough

A p-value can address whether the observed data are inconsistent with a specified null hypothesis, but it does not describe the magnitude of the R0 difference. A very small p-value can accompany a modest treatment difference in a large study, while a clinically substantial difference can produce an imprecise p-value in a smaller study.

For a phase 2 randomized trial with 130 enrolled participants, the statistical interpretation should therefore consider the estimated R0 rates, the absolute or relative effect measure, the confidence interval, and the definition of the analysis population together.

19. Planned Analysis: A Practical Statistical Workflow

A rigorous analysis would begin by defining the pathological R0 outcome and the analysis population. The R0 rate would then be estimated separately for the two randomized strategies. The treatment effect could be expressed using an appropriate absolute or relative measure, followed by a confidence interval and a prespecified hypothesis test where applicable.

The statistical analysis should also preserve the distinction between the randomized treatment comparison and any post-randomization subset analysis. This distinction is particularly important when the endpoint requires patients to reach surgery and provide an assessable pathological specimen.

20. What Can Be Learned From the Trial Design

21. Important Statistical Limitations

• No numerical efficacy result in the posted analyses: the registry identifies the primary endpoint but does not post an estimate, confidence interval, or p-value for it.
• No reported arm-level R0 rates: the available record does not provide the numerical R0 rate for either randomized strategy.
• No reported statistical test: the registry does not identify the formal hypothesis test used for the primary endpoint.
• No reported effect measure: the registry does not specify whether the primary treatment comparison was expressed as a risk difference, risk ratio, odds ratio, or another measure.
• No reported confidence-interval method: the record does not identify the interval-estimation procedure for the primary endpoint.
• No reported multiplicity strategy: the available record does not describe multiplicity adjustment or an endpoint hierarchy.
• No reported stratification: the available record does not identify randomization strata or stratified statistical analyses.
• No reported interim-analysis framework: the available record does not describe interim efficacy or futility monitoring.

22. Why This Trial Is a Useful Biostatistics Example

PRODIGE 44 demonstrates how clinical-trial statistics can remain relatively straightforward at the endpoint level while still requiring careful interpretation. The trial is randomized and parallel, but its registered primary outcome is not a survival endpoint. It is a histological R0 resection margin rate assessed over a defined period.

That distinction matters. Applying a familiar survival-analysis framework simply because the trial concerns cancer would be inappropriate for the primary endpoint as registered. Instead, the statistical analysis should begin with the binary structure of the R0 outcome, define the appropriate analysis population, estimate the treatment-group proportions, and report an effect measure with its uncertainty.

The trial also illustrates why statistical interpretation cannot be separated from endpoint definition. The same numerical percentage can have a different meaning depending on whether it is calculated among all randomized participants, among patients undergoing surgery, or among another prespecified population.

23. Record Summary

PRODIGE 44 is a completed randomized phase 2 trial of neoadjuvant mFolfirinox with or without preoperative concomitant chemoradiotherapy in patients with borderline resectable pancreatic carcinoma. The study enrolled 130 participants across two parallel, unmasked treatment arms. Its registered primary endpoint is the histological R0 resection margin rate, assessed over a period of up to 7.5 months.

From a statistical perspective, the primary endpoint is a binary surgical-pathology outcome and is therefore naturally approached through a comparison of proportions rather than a time-to-event model. The most important interpretive elements are the definition of the analysis population, the R0 rate in each randomized group, the selected treatment-effect measure, its confidence interval, and the corresponding hypothesis-testing framework.

24. Sources

• ClinicalTrials.gov: PRODIGE 44 (NCT02676349).