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
FRESCO was a randomized, parallel-group phase 3 trial evaluating fruquintinib versus placebo in patients with colorectal cancer. The registry records an enrollment of 416 participants across 2 arms and identifies overall survival as the primary endpoint.
| Feature | FRESCO |
|---|---|
| Trial acronym | FRESCO |
| NCT identifier | NCT02314819 |
| Phase | Phase 3 |
| Status | Completed |
| Condition | Colorectal Cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 416 |
| Interventions | Fruquintinib; placebo |
| Lead sponsor | Hutchison Medipharma Limited |
| Sponsor type | Industry |
| Start date | 2014-12 |
| Primary completion date | 2017-01 |
2. Clinical Question
The statistical question is whether randomized assignment to fruquintinib, compared with randomized assignment to placebo, is associated with a difference in overall survival among patients with colorectal cancer enrolled in this phase 3 treatment trial.
Population
Patients with colorectal cancer enrolled in the FRESCO phase 3 trial.
Intervention
Fruquintinib.
Comparator
Placebo.
Primary question
What is the difference in overall survival between the randomized treatment groups?
3. Trial Design
Fruquintinib
- Drug intervention
- Randomized treatment arm
- Evaluated for overall survival
Placebo
- Placebo intervention
- Randomized comparator arm
- Evaluated for overall survival
The combination of randomization, parallel assignment, and quadruple masking defines the principal structural features of the comparison. Randomization is particularly important statistically because it establishes treatment assignment before the outcome is observed, allowing the between-group comparison to be interpreted within the framework of a randomized experiment.
4. Endpoints
| Endpoint | Registry definition | Time frame | Assessment |
|---|---|---|---|
| Overall survival | From randomization until death due to any cause | Assessed up to 2 year | Every two months after end of treatment observation period at 30 days after the last medication |
Overall survival is a time-to-event endpoint. The event is death due to any cause, and the analysis begins at randomization. The registry describes assessments every two months after the end of the treatment observation period, which occurs at 30 days after the last medication.
5. Statistical Methodology
The primary endpoint is overall survival, so the natural statistical framework is a time-to-event analysis rather than a simple comparison of proportions at a single fixed time point. The registry identifies the endpoint and its time frame but does not report a posted statistical analysis specifying a particular survival model or hypothesis-testing procedure.
Time-to-event analysis
For overall survival, each participant contributes a follow-up time measured from randomization. If death occurs, the observation ends with the event. If a participant remains alive at the last relevant assessment, the observation may be right-censored at that point in a conventional survival analysis.
Here, S(t) represents the probability of surviving beyond time t, where T is the time from randomization to death.
Kaplan-Meier estimation
A standard descriptive analysis for overall survival would use the Kaplan-Meier estimator to estimate the survival function in each randomized treatment group. This method allows patients who have not experienced death by the end of their observed follow-up to contribute information without being treated as if death had occurred.
At each observed event time, di is the number of deaths and ni is the number of participants at risk immediately before that time.
Hazard ratio as a comparative measure
A common inferential measure for randomized overall-survival comparisons is the hazard ratio, often estimated with a Cox proportional-hazards model. If a hazard ratio were reported, it would summarize the relative instantaneous event rate between the randomized groups under the model.
An HR below 1 would correspond to a lower estimated instantaneous rate of death in the fruquintinib group under the fitted model. An HR of 1 would indicate no estimated relative difference in the modeled hazard.
Log-rank comparison
A log-rank test is a conventional way to compare survival distributions between randomized groups. It uses the ordering of observed event times and compares the observed number of events in each group with the numbers expected under a common survival distribution.
The registry record does not report that a log-rank test, Cox model, Kaplan-Meier estimator, or another specific inferential procedure was used. These methods therefore describe the standard statistical framework appropriate for an overall-survival endpoint rather than a reported FRESCO analysis.
6. Planned Analysis
The registry identifies overall survival as the primary endpoint and defines it as time from randomization until death due to any cause, assessed up to 2 year. Results and formal statistical analyses have not been posted on ClinicalTrials.gov.
Primary endpoint
Time from randomization until death due to any cause.
Assessment: up to 2 year, with assessments every two months after the end of the treatment observation period at 30 days after the last medication.
For an endpoint of this type, a complete statistical analysis would ordinarily describe the survival distribution in each randomized arm and provide a between-group comparison. Kaplan-Meier curves would provide a nonparametric estimate of survival over time. A log-rank framework could provide a formal comparison of survival distributions, while a Cox proportional-hazards model could provide a hazard ratio and confidence interval if that model were prespecified and used.
The important statistical distinction is between the endpoint itself and the method used to analyze it. "Overall survival" defines what is measured; it does not, by itself, specify whether the final comparison should be expressed as a hazard ratio, a difference in survival probability at a fixed time, a difference in restricted mean survival time, or another estimand.
7. Statistical Methods Explained
What makes overall survival a time-to-event endpoint?
Overall survival records not only whether death occurs, but also when it occurs relative to randomization. A participant alive at the end of observed follow-up does not provide the same information as a participant who died early. Survival methods preserve this timing information and account for censoring.
Why is randomization important for the survival comparison?
Randomization assigns participants to treatment groups before the outcome is observed. In a properly conducted randomized comparison, this reduces systematic differences in both measured and unmeasured characteristics between groups in expectation. The subsequent difference in survival can therefore be evaluated in relation to randomized treatment assignment rather than relying solely on observational adjustment.
What does a hazard ratio measure?
A hazard ratio compares the instantaneous event rates between two groups under a time-to-event model. It is a relative measure and should not be interpreted as the percentage of patients who survive, the percentage of patients cured, or a direct difference in median survival.
Why is a confidence interval important?
A confidence interval communicates statistical uncertainty around an estimated treatment effect. A narrow interval indicates greater precision than a wide interval, all else being equal. The interval is not a range containing the individual treatment effects experienced by patients.
Why does censoring matter?
Survival analysis commonly encounters participants who are alive when observation ends or who have no event recorded after their last assessment. These observations are typically censored rather than classified as deaths. The validity of a survival analysis depends on appropriate handling of censoring and on assumptions connecting censoring with the event process.
What is the proportional-hazards assumption?
A Cox proportional-hazards model assumes that the relative hazard between treatment groups is stable over time in the relevant sense. If the treatment effect changes substantially over follow-up, a single hazard ratio can compress a more complicated time-varying pattern into one summary measure. The registry does not report whether proportional hazards were assessed or whether a Cox model was used.
8. Interpreting the Primary Endpoint
Overall survival is a direct time-to-death endpoint beginning at randomization. It incorporates deaths from any cause rather than restricting the event definition to disease progression or a particular cause of death.
The endpoint definition alone does not establish the magnitude or direction of a treatment effect. Those conclusions require the observed event data and an appropriate statistical comparison between the randomized groups.
The registry specifies assessment up to 2 year. A survival estimate at a particular time point is therefore a different quantity from the complete survival distribution and should not be treated as interchangeable with a median survival or a hazard ratio.
9. Censoring and Follow-Up
Time-to-event trials frequently contain incomplete event histories. A participant may remain alive when the observation period ends, for example, and therefore contribute survival information up to the last time at which survival status is known. This is fundamentally different from treating the participant as having experienced the event.
Event
For FRESCO's primary endpoint, the event is death due to any cause.
Time origin
The registry defines follow-up from randomization.
Assessment schedule
The registry specifies assessment every two months after the end of treatment observation.
Observation period
The registry describes the end of treatment observation as 30 days after the last medication.
These details matter because survival analysis is built around the exact relationship between the time origin, event definition, follow-up process, and censoring mechanism. Changing any of those elements can change the estimand being analyzed.
10. Randomization and Masking
The FRESCO registry record describes the study as randomized, parallel, and quadruple-masked. These design features are relevant to statistical interpretation because they concern how treatment assignment was generated and how knowledge of treatment assignment was controlled during the study.
| Design feature | Registry description | Statistical relevance |
|---|---|---|
| Allocation | Randomized | Creates the basis for a causal comparison between randomized groups. |
| Design model | Parallel | Participants are assigned to treatment groups that are compared over follow-up. |
| Masking | Quadruple | Limits knowledge of treatment assignment among the groups covered by the masking scheme. |
| Primary purpose | Treatment | The study is designed to evaluate a treatment intervention. |
Randomization and masking address different sources of potential bias. Randomization concerns treatment assignment; masking concerns knowledge of assignment. Neither feature, considered alone, determines the appropriate survival estimand or guarantees that every statistical assumption will hold.
11. What the Sample Size Represents
The registry records an enrollment of 416 participants and 2 treatment arms. Enrollment is the number of participants entered into the study, but it is not itself a measure of statistical significance, treatment effect, or precision.
For an overall-survival analysis, the number and timing of deaths are central to the amount of statistical information available for estimating a treatment effect. The registry information presented here does not provide an event count or a posted power calculation.
This distinction is especially important for survival trials. Two studies with the same number of enrolled participants can provide different amounts of statistical information if their event rates, follow-up durations, or censoring patterns differ.
12. Why a Single P-Value Would Not Be Enough
If a formal survival comparison produces a p-value, that value addresses a hypothesis-testing question under the specified statistical model. It does not describe the magnitude of the treatment effect or its clinical importance.
Effect estimate
A hazard ratio, survival probability difference, or another prespecified estimand describes the estimated treatment effect.
Confidence interval
The confidence interval communicates uncertainty around the estimated effect.
P-value
The p-value evaluates compatibility with a specified null hypothesis under the statistical model.
Clinical meaning
Clinical interpretation requires understanding the magnitude, time horizon, precision, and context of the effect rather than relying on a p-value alone.
Because the ClinicalTrials.gov record does not post a formal statistical analysis for FRESCO's primary endpoint, there is no registry-reported p-value or effect estimate to interpret here.
13. Common Statistical Pitfalls for Overall Survival
- Confusing hazard with probability: a hazard ratio is not the same as a risk ratio or an absolute survival difference.
- Ignoring censoring: participants who are alive at the end of follow-up should not simply be counted as if they had experienced the event.
- Equating the median with the entire survival distribution: the median summarizes one point of the survival curve and does not describe the full trajectory.
- Assuming proportional hazards without assessment: a Cox hazard ratio relies on a model whose interpretation can become problematic when relative hazards vary substantially over time.
- Using a fixed-time survival percentage as the only result: a two-year survival probability answers a different question from the entire time-to-event comparison.
- Interpreting enrollment as information: the statistical information available for a survival comparison depends importantly on observed events and follow-up.
- Confusing statistical significance with effect size: a p-value does not tell readers how large an effect is.
14. Registry Timeline
Trial start
The registry records the FRESCO study as starting in December 2014.
Randomized treatment evaluation
The study is registered as a phase 3, randomized, parallel-group treatment trial with 2 arms.
Enrollment
The registry records enrollment of 416 participants.
Primary completion
The registry records a primary completion date of January 2017.
Trial status
The ClinicalTrials.gov record lists the study status as completed.
15. Limitations
- No posted statistical analysis: the registry identifies overall survival as the primary endpoint but does not post a formal statistical analysis with an effect estimate, confidence interval, or p-value.
- Limited endpoint information: the registered primary endpoint is overall survival. Additional efficacy or safety endpoints are not specified in the information represented here.
- No event counts: the registry information does not provide the number of deaths contributing to the primary analysis.
- No reported analysis model: the registry information does not identify a particular survival model or hypothesis-testing procedure for the primary endpoint.
- No reported confidence interval: without a posted analysis, the precision of an estimated treatment effect cannot be assessed from the registry record.
- No reported p-value: the registry information does not provide a formal hypothesis-test result for overall survival.
- No reported subgroup analyses: the available record does not provide subgroup-specific treatment effects for the primary endpoint.
- No reported baseline analysis: the available record does not provide a baseline characteristics table for the randomized groups.
These limitations affect what can be concluded statistically from the registry record. They do not change the registered design or endpoint definition; they determine how much of the statistical evidence can be evaluated from the information posted there.
16. Why This Trial Matters Statistically
FRESCO provides a useful teaching example of how a randomized oncology trial can be understood from its statistical architecture even when the registry record does not contain a posted results analysis. The key elements are the randomized comparison, the parallel treatment structure, the masking specification, and the use of overall survival as the primary time-to-event endpoint.
| Concept | How it appears in FRESCO |
|---|---|
| Randomization | The registry identifies randomized allocation. |
| Parallel design | The study has 2 parallel treatment arms. |
| Masking | The registry identifies quadruple masking. |
| Time-to-event endpoint | Overall survival is measured from randomization until death due to any cause. |
| Kaplan-Meier estimation | A conventional descriptive method for the registered overall-survival endpoint. |
| Hazard ratio | A conventional relative treatment-effect measure for a Cox survival analysis. |
| Confidence interval | Would quantify uncertainty around a reported survival effect estimate. |
| Log-rank testing | A conventional method for comparing survival distributions between randomized groups. |
| Censoring | Important whenever participants remain alive at the end of observed follow-up. |
| Analysis population | The interpretation of a survival result depends on the population defined for the prespecified analysis. |
17. A Practical Framework for Reading the FRESCO Endpoint
A statistically complete reading of an overall-survival endpoint should move through several questions in sequence.
1. What is measured?
Time from randomization until death due to any cause.
2. Who is compared?
Participants randomized to fruquintinib versus participants randomized to placebo.
3. What is the estimand?
The registry defines the endpoint but does not report a specific statistical estimand beyond overall survival.
4. What analysis is used?
The registry does not report a posted statistical method for the primary endpoint.
5. How precise is the result?
A confidence interval would normally provide information about statistical uncertainty, but none is posted in the available registry analysis.
6. What does the result mean?
The treatment effect should be interpreted through the prespecified estimand, analysis population, follow-up, censoring rules, and uncertainty interval.
18. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The registered primary endpoint is a time-to-event outcome beginning at randomization. A formal statistical comparison would ordinarily estimate and compare survival distributions while accounting for incomplete follow-up.
Clinical interpretation
The clinical meaning of a treatment comparison depends on the magnitude and durability of any observed difference in overall survival, its precision, and the context of treatment and follow-up.
The two perspectives should remain distinct. Statistical analysis describes the evidence for a difference between randomized groups; clinical interpretation considers what the size and pattern of that difference mean in the context of patient outcomes.
19. Sources
- ClinicalTrials.gov: NCT02314819.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through the Clinical Biostats knowledge graph
Explore the broader statistical framework behind randomized clinical trials, survival endpoints, censoring, and treatment-effect estimation.
20. Record Summary
FRESCO is registered as a completed, randomized, parallel-group phase 3 treatment trial in colorectal cancer, with 416 participants assigned across 2 arms evaluating fruquintinib and placebo. Its registered primary endpoint is overall survival, defined from randomization until death due to any cause and assessed up to 2 year, with assessments every two months after the end of the treatment observation period at 30 days after the last medication.
From a statistical perspective, the central feature of FRESCO is therefore a randomized time-to-event comparison. Kaplan-Meier estimation, log-rank testing, and Cox proportional-hazards modeling are standard approaches that can be used to describe and compare overall survival, but the ClinicalTrials.gov record does not post a formal statistical analysis specifying which of these methods was used. No effect estimate, confidence interval, p-value, event count, subgroup result, or formal safety analysis is reported in the registry information represented here.