This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. The numerical results on this page are restricted to the trial data reported for FRESCO-2.
1. Trial at a Glance
FRESCO-2 was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating fruquintinib plus best supportive care against placebo plus best supportive care in participants with metastatic colorectal cancer. The registry reports 691 enrolled participants and a primary endpoint of overall survival.
| Feature | FRESCO-2 |
|---|---|
| Phase | Phase 3 |
| Condition | Metastatic colorectal cancer; metastatic colon cancer |
| Design | Randomized, parallel-group, quadruple-masked |
| Primary purpose | Treatment |
| Allocation | Randomized |
| Enrollment | 691 |
| Arms | 2 |
| Interventions | Fruquintinib; placebo |
| Primary endpoint | Overall Survival (OS) |
| Primary endpoint type | Time-to-event |
| Trial status | Completed |
| Start | August 12, 2020 |
| Primary completion | July 29, 2022 |
| Lead sponsor | Hutchison Medipharma Limited |
2. Clinical Question
The primary statistical question was whether participants randomized to fruquintinib plus best supportive care experienced a different overall-survival distribution from participants randomized to placebo plus best supportive care. The registered hypothesis type for the primary analysis was superiority.
Population
Participants with metastatic colorectal cancer; the registry also lists metastatic colon cancer as a condition.
Intervention
Fruquintinib plus best supportive care.
Comparator
Placebo plus best supportive care.
Primary question
Does fruquintinib plus best supportive care improve overall survival relative to placebo plus best supportive care?
3. Trial Design
Fruquintinib + BSC
- Fruquintinib
- Best supportive care
- Randomized treatment group
Placebo + BSC
- Placebo
- Best supportive care
- Randomized comparator group
The design is explicitly classified as randomized, parallel, and quadruple-masked. These features matter statistically because randomization establishes the treatment comparison, the parallel structure creates two concurrently followed groups, and masking can reduce opportunities for knowledge of assignment to influence treatment administration or assessment.
4. Endpoints
The registered primary endpoint was a single time-to-event endpoint: overall survival. The registry also reports secondary efficacy and exposure-response analyses.
| Endpoint | Registry definition / time frame | Type |
|---|---|---|
| Overall Survival (OS) | From date of randomization to death from any cause (up to 22 months). OS was defined as the time in months from randomization to death from any cause. Participants without report of death at the time of analysis were censored at the date last known alive. | Time-to-event |
| Progression Free Survival (PFS) | From randomization until the first documentation of objective progression or death, whichever comes first (up to 22 months) | Time-to-event |
| Objective Response Rate (ORR) | From randomization until the first documentation of best overall response (up to 22 months) | Binary |
| Disease Control Rate (DCR) | From randomization until the first documentation of best overall response (up to 22 months) | Binary |
The registry definition gives the OS calculation explicitly as (date of death or last known alive − date of randomization + 1) / 30.4375. This makes the endpoint a duration measured in months while preserving censoring for participants who were alive at the analysis.
5. Statistical Methodology
Intention-to-treat analysis
The primary OS analysis used the ITT population, defined in the registry analysis as including all randomized participants. This is an important design principle: the primary efficacy comparison follows randomized assignment rather than redefining the groups according to treatment exposure after randomization.
Stratified log-rank test
The reported primary method was the stratified log-rank test. The registry analysis states that the hazard ratio and its confidence interval were obtained from a stratified Cox proportional-hazards model accounting for the randomization schedule stratification factors.
The time-to-event framework incorporates both the timing of deaths and the information contributed by participants who had not experienced death when the analysis was performed.
Stratified Cox proportional-hazards model
The primary effect measure was a stratified hazard ratio. The registry states that the hazard ratio was calculated using a stratified Cox proportional-hazards model accounting for the randomization schedule stratification factors.
For the primary OS result, the estimated hazard ratio was 0.662. A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the fruquintinib group relative to the placebo group under the fitted model. A value of 0.662 corresponds to an estimated hazard approximately 33.8% lower than the comparator because \(1-0.662=0.338\). This is a mathematical interpretation of the reported estimate, not a statement that 33.8% of participants avoided death.
Cochran-Mantel-Haenszel test
For ORR and DCR, the registry reports the Cochran-Mantel-Haenszel test. This method allows a categorical comparison while accounting for the randomization schedule stratification factors. The reported adjusted differences and their 95% confidence intervals were calculated using the Wald method from the Cochran-Mantel-Haenszel test.
Wald / z-test
The registry also reports Wald tests for the exposure-response analyses. These analyses are distinct from the primary randomized treatment comparison: they evaluate relationships between exposure measures such as CminSS or CmaxSS and efficacy or safety parameters in pooled fruquintinib populations.
6. Results: Overall Survival
Overall survival was the single registered primary endpoint. The analysis included all randomized participants in the ITT population and compared the fruquintinib plus BSC group with the placebo plus BSC group.
Primary endpoint: Overall Survival
95% CI: 0.549–0.800 · P < .001
Stratified log-rank test; stratified Cox proportional-hazards model
| Primary endpoint | Fruquintinib + BSC vs placebo + BSC |
|---|---|
| Analysis population | ITT population; all randomized participants |
| Endpoint | Overall Survival (OS) |
| Time frame | From date of randomization to death from any cause (up to 22 months) |
| Method | Stratified log-rank test |
| Effect measure | Stratified Hazard Ratio |
| Estimate | 0.662 |
| 95% CI | 0.549–0.800 |
| P-value | < .001 |
| Hypothesis | Superiority |
What the estimate means: The reported HR of 0.662 means that, under the stratified Cox model, the estimated instantaneous hazard of death in the fruquintinib plus BSC group was 0.662 times the estimated hazard in the placebo plus BSC group. Equivalently, the estimated hazard was approximately 33.8% lower.
What it does not mean: A hazard ratio of 0.662 is not a 33.8% absolute reduction in mortality, does not mean that 33.8% of participants benefited, and does not imply that each individual participant experienced the same relative reduction.
What the confidence interval says: The 95% CI of 0.549–0.800 describes the statistical uncertainty around the estimated hazard ratio under the model and analysis framework. It does not describe the range of effects among individual patients.
Why the p-value is different: The P < .001 result addresses the evidence against the null hypothesis in the specified superiority test. It does not quantify the size or clinical importance of the treatment effect. The effect size is described by the hazard ratio and its confidence interval.
Important time-to-event caution: The hazard ratio is a model-based summary and relies on the Cox-model framework. The ClinicalTrials.gov record does not provide a formal assessment of the proportional-hazards assumption, so the single HR should not be treated as a complete description of how treatment effects may have varied over time.
How to read the primary result
The most direct statistical conclusion from the ClinicalTrials.gov record is that the prespecified primary OS comparison produced an estimated hazard ratio below 1, a 95% confidence interval entirely below 1, and a P-value below .001. Together, these quantities describe both the direction of the estimated treatment effect and the statistical evidence for superiority.
The page does not report a median OS, survival probabilities at selected time points, or numbers of deaths by treatment arm because those quantities are not included in the ClinicalTrials.gov record. A hazard ratio alone should not be converted into an absolute survival benefit without the underlying survival distribution.
7. Secondary Results: Progression-Free Survival
PFS was a secondary time-to-event endpoint defined from randomization until the first documentation of objective progression or death, whichever came first. The analysis used the ITT population and a stratified log-rank test.
Progression-Free Survival
95% CI: 0.267–0.386 · P < .001
Stratified log-rank test; stratified Cox proportional-hazards model
| Endpoint | Estimate / result |
|---|---|
| Analysis population | ITT population; all randomized participants |
| Time frame | From randomization until first documentation of objective progression or death, whichever comes first (up to 22 months) |
| Method | Stratified log-rank test |
| Hazard ratio | 0.321 |
| 95% CI | 0.267–0.386 |
| P-value | < .001 |
| Hypothesis | Superiority |
The PFS HR of 0.321 indicates an estimated instantaneous rate of progression or death approximately 67.9% lower in the fruquintinib group under the reported Cox model. The calculation \(1-0.321=0.679\) expresses the relative hazard difference implied by the estimate.
The 95% CI of 0.267–0.386 indicates uncertainty around that estimated relative effect. It does not indicate that the true effect for individual participants lies somewhere inside that interval.
The P-value of < .001 addresses the statistical comparison under the reported test; it does not tell us that the treatment effect is "67.9% clinically important," nor does it provide an absolute difference in progression-free survival.
The ClinicalTrials.gov record does not report median PFS, landmark PFS rates, numbers of progression events, or a Kaplan-Meier curve. Those quantities are therefore not presented here.
8. Secondary Results: Objective Response Rate
ORR was defined as the first documentation of best overall response during the period from randomization until up to 22 months. The registry identifies the endpoint as binary and reports a Cochran-Mantel-Haenszel analysis in the ITT population.
Objective Response Rate
95% CI: 0.4–2.7 · P = .059
Cochran-Mantel-Haenszel test with Wald confidence interval
| Endpoint | Reported result |
|---|---|
| Analysis population | ITT population; all randomized participants |
| Time frame | From randomization until first documentation of best overall response (up to 22 months) |
| Method | Cochran-Mantel-Haenszel |
| Effect measure | Adjusted difference |
| Estimate | 1.5 |
| 95% CI | 0.4–2.7 |
| P-value | .059 |
| Hypothesis | Superiority |
The registry reports the adjusted difference in the endpoint's percentage-of-participants scale. Because the ClinicalTrials.gov record does not identify the underlying response percentages for the two randomized groups, this page does not reconstruct or infer those percentages from the adjusted difference.
9. Secondary Results: Disease Control Rate
DCR was another binary endpoint measured from randomization until the first documentation of best overall response, with a time frame of up to 22 months. The analysis used the Cochran-Mantel-Haenszel method in the ITT population.
Disease Control Rate
95% CI: 32.8–46.0 · P < .001
Cochran-Mantel-Haenszel test with Wald confidence interval
The reported adjusted difference of 39.4 is expressed on the percentage-of-participants scale used for this binary endpoint. The 95% CI of 32.8–46.0 quantifies uncertainty around that adjusted difference, while P < .001 provides the statistical evidence associated with the reported superiority test.
The result should not be translated into an unadjusted between-arm percentage difference unless those underlying arm-specific response rates are explicitly available. The ClinicalTrials.gov record reports the adjusted difference, not the two raw DCR percentages.
10. Exposure-Response Analyses
The registry contains additional secondary analyses examining relationships between fruquintinib exposure and efficacy or safety parameters. These analyses are structurally different from the randomized comparison because they use pooled fruquintinib populations and exposure measures rather than comparing the two randomized treatment arms.
Exposure and Overall Survival
| Exposure-response analysis | Method | P-value |
|---|---|---|
| CminSS based on the starting dose for OS exposure-response analyses | Wald test | 0.0600 |
| CminSS based on the adjusted RDI for OS exposure-response analyses | Wald test | 0.8065 |
The registry describes these analyses as including participants evaluable for population PK analysis who had PK parameter estimates enabling estimation of fruquintinib exposure and efficacy relationships. The reported outcome unit is 1/(ng/mL).
Exposure and Safety
| Exposure-response analysis | Method | P-value |
|---|---|---|
| CmaxSS for any-grade dermatological toxicity | Wald test | 0.265 |
| CmaxSS for grade 3+ dermatological toxicity | Wald test | 0.0159 |
| CmaxSS for any-grade proteinuria | Wald test | 0.484 |
| CmaxSS for grade 3+ proteinuria | Wald test | 0.642 |
| CmaxSS for grade hemorrhage | Wald test | 0.166 |
11. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized arm. The registry identifies 173/456 affected participants among the Fruquintinib + BSC group and 88/230 among the Placebo + BSC group.
| Safety measure | Fruquintinib + BSC | Placebo + BSC |
|---|---|---|
| Serious adverse events | 173/456 | 88/230 |
The numerator and denominator are reported exactly as reported in the registry. This page does not convert them into percentages because the trial data request that numerical values not be recomputed.
Safety is a separate statistical domain from efficacy. The serious-adverse-event counts describe the occurrence of serious adverse events among the reported at-risk populations; they do not establish whether any individual event was caused by treatment, nor do they summarize the full adverse-event profile.
The ClinicalTrials.gov record does not provide a complete table of adverse-event types, grades, treatment discontinuations, deaths due to adverse events, or exposure-adjusted safety rates. Those quantities are therefore not inferred.
12. Statistical Methods Explained
Why was a stratified log-rank test used for overall survival?
Overall survival is a time-to-event endpoint because both the occurrence and timing of death matter, while some participants may remain alive at the analysis and therefore contribute censored follow-up. A log-rank framework compares the survival experience between randomized groups across follow-up. The registry specifies a stratified log-rank test, aligning the comparison with the randomization schedule stratification structure.
What does an OS hazard ratio of 0.662 mean?
A hazard ratio of 0.662 means that the fitted model estimates the instantaneous hazard of death in the fruquintinib group to be 0.662 times that in the placebo group. The corresponding relative hazard reduction implied by the estimate is approximately 33.8%. This does not mean that survival time is 33.8% longer, that mortality is reduced by 33.8 percentage points, or that every participant experiences the same effect.
Why is the confidence interval important?
The estimate alone does not describe statistical precision. For OS, the 95% CI is 0.549–0.800. It communicates uncertainty around the hazard-ratio estimate under the specified analysis framework. A confidence interval also helps distinguish a precise estimate from one that is highly uncertain; the width of the interval is therefore important information alongside the point estimate.
Why doesn't the P-value measure effect size?
The P-value answers a hypothesis-testing question under the specified statistical model and null hypothesis. It does not measure how large the treatment effect is. In FRESCO-2, the primary OS P-value is < .001, while the estimated effect is separately summarized by the HR of 0.662 and its 95% CI of 0.549–0.800.
Why use the Cochran-Mantel-Haenszel test for ORR and DCR?
ORR and DCR are binary outcomes rather than time-to-event outcomes. The registry reports the Cochran-Mantel-Haenszel method, which provides a stratified comparison while accounting for the randomization schedule stratification factors. The associated adjusted difference is then accompanied by a Wald confidence interval.
Why is ITT important in this trial?
The primary efficacy analysis included all randomized participants. An ITT framework maintains the treatment assignment established by randomization rather than restricting the efficacy comparison to participants who completed treatment or adhered perfectly to the assigned intervention. That distinction is particularly important because post-randomization treatment exposure can differ from randomized assignment.
How are the exposure-response P-values different from the primary OS P-value?
The primary OS P-value comes from the randomized comparison of fruquintinib plus BSC versus placebo plus BSC. The exposure-response analyses instead examine relationships between estimated drug exposure and efficacy or safety parameters within pooled fruquintinib populations. A significant or nonsignificant exposure-response P-value therefore answers a different question from the primary randomized treatment comparison.
13. Primary Endpoint: Statistical Interpretation in Context
Relative effect
The primary OS HR of 0.662 summarizes the relative treatment effect on the hazard scale under the stratified Cox model.
Statistical evidence
The two-sided 95% CI is 0.549–0.800 and the reported P-value is < .001.
Absolute effect
The ClinicalTrials.gov record does not include median OS or landmark survival probabilities, so an absolute OS benefit cannot be reconstructed here.
Analysis population
The primary efficacy analysis was based on the ITT population, including all randomized participants.
This combination of information is more informative than the P-value alone. The hazard ratio provides the estimated relative effect, the confidence interval describes uncertainty, and the analysis population identifies the population to which the randomized comparison applies.
At the same time, a hazard ratio is not a substitute for the survival curve. Because the ClinicalTrials.gov record does not include the underlying event and censoring times, this page does not attempt to construct a Kaplan-Meier curve or calculate medians that are not reported in the registry data.
14. Understanding the Difference Between OS, PFS, ORR, and DCR
| Endpoint | Question answered | Statistical framework in FRESCO-2 |
|---|---|---|
| OS | How long participants remain alive after randomization | Stratified log-rank; stratified Cox HR |
| PFS | How long participants remain free of objective progression or death | Stratified log-rank; stratified Cox HR |
| ORR | Whether a participant achieves the defined best overall response | Cochran-Mantel-Haenszel; adjusted difference |
| DCR | Whether a participant achieves disease control according to the endpoint definition | Cochran-Mantel-Haenszel; adjusted difference |
These endpoints should not be treated as interchangeable. OS is a direct time-to-death endpoint. PFS incorporates progression and death. ORR and DCR are categorical measures based on response assessment. Consequently, a hazard ratio for OS cannot be directly compared numerically with an adjusted percentage difference for ORR or DCR.
15. Censoring and the Meaning of “Up to 22 Months”
The primary OS endpoint is defined from randomization to death from any cause, with participants without a report of death censored at the date last known alive. This is the standard structure of a right-censored time-to-event endpoint.
Censoring allows participants who remain alive at the analysis to contribute information without treating the analysis date as if it were their death date.
The phrase up to 22 months is the registry time frame for the primary endpoint. It should not be interpreted as saying that every participant was followed for exactly 22 months. The ClinicalTrials.gov record does not provide the individual follow-up distribution.
16. Analysis Population and Generalizability
The ITT definition is explicit for the primary and several secondary analyses: all randomized participants are included. This is important because the treatment comparison is anchored to randomization rather than to a post-randomization selection process.
At the same time, the exposure-response analyses use a different population concept. The registry text states that efficacy exposure-response analyses included participants evaluable for population PK analysis and therefore having PK parameter estimates that enabled estimation of fruquintinib exposure. Safety exposure-response analyses use an analogous population- and PK-evaluable framework.
17. Multiplicity, Interim Analysis, and Other Design Issues
The ClinicalTrials.gov record identifies a single registered primary endpoint and report a superiority hypothesis for that endpoint. They also report multiple secondary analyses, including PFS, ORR, DCR, and exposure-response analyses.
| Design topic | What the ClinicalTrials.gov record establishes |
|---|---|
| Primary endpoint | Overall Survival (OS) |
| Primary hypothesis | Superiority |
| Primary analysis | Stratified log-rank test with stratified Cox hazard ratio |
| Secondary endpoints | PFS, ORR, DCR, and exposure-response analyses |
| Multiplicity procedure | Not reported in the ClinicalTrials.gov record |
| Interim analysis | Not reported in the ClinicalTrials.gov record |
| Missing-data / imputation strategy | Not reported in the ClinicalTrials.gov record |
| Non-inferiority margin | Not applicable to the reported superiority hypothesis; no non-inferiority margin is reported |
| Crossover | Not reported in the ClinicalTrials.gov record |
| Bayesian methods | Not reported in the ClinicalTrials.gov record |
These omissions are important. Multiple reported P-values do not automatically constitute a single multiplicity-adjusted family, and the ClinicalTrials.gov record does not establish how any such family was handled. Likewise, the presence of several statistical analyses does not establish that they were all confirmatory.
18. What the Primary P-value Does and Does Not Tell Us
The primary OS result reports P < .001. This is evidence against the null hypothesis specified for the superiority analysis under the reported stratified log-rank framework.
The P-value does not tell us that the treatment effect is large, that the effect is clinically meaningful, or that the probability of treatment benefit for an individual participant is greater than a particular percentage. Those questions require effect estimates, uncertainty intervals, and clinical context.
A frequent statistical misconception is to interpret a P-value as the probability that the null hypothesis is true. That is not what the reported P-value represents. It is calculated under a specified null model and measures the extremeness of the observed data relative to that model.
19. Hazard Ratio: A Deeper Interpretation
The primary HR of 0.662 is a relative measure of the event hazard. In a Cox model, the hazard represents an instantaneous event rate conditional on having remained event-free up to a particular time, subject to the model framework.
The corresponding relative hazard reduction implied by the estimate is approximately 33.8%.
Several common interpretations should be avoided. The HR is not an absolute risk reduction, does not equal the ratio of median survival times, and does not mean that every participant's individual probability of death is reduced by exactly the same percentage.
The PFS HR of 0.321 has the same mathematical structure, but it applies to the composite time-to-event endpoint of objective progression or death. The two HRs therefore cannot be interpreted as if they measure the same event.
20. Limitations
- Registry-level reporting: the ClinicalTrials.gov record contains selected registry results and methods rather than a complete statistical analysis plan or full clinical study report.
- Missing absolute survival summaries: median OS, landmark survival probabilities, event counts, and Kaplan-Meier estimates are not reported, so the absolute magnitude of the OS effect cannot be reconstructed from the reported HR alone.
- Missing PFS summaries: median PFS, event counts, and landmark PFS estimates are not included in the ClinicalTrials.gov record.
- Incomplete response data: the registry provides adjusted differences for ORR and DCR but not the corresponding raw response percentages by randomized arm.
- Multiplicity: the ClinicalTrials.gov record does not describe a multiplicity-control procedure for the primary and secondary analyses. Individual P-values therefore should be interpreted in the context of their specific analyses rather than automatically treated as independent confirmatory tests.
- Proportional hazards: the primary HR comes from a Cox model, but the ClinicalTrials.gov record does not report a formal assessment of the proportional-hazards assumption.
- Exposure-response selection: PK exposure-response analyses use evaluable, pooled fruquintinib populations and therefore do not have the same causal structure as the randomized ITT comparison.
- Safety detail: only serious adverse-event counts by arm are reported here. The complete safety profile cannot be characterized from those counts alone.
- Unreported design features: details concerning interim analysis, missing-data handling, crossover, and Bayesian methods are not contained in the ClinicalTrials.gov record and are not inferred.
21. Why This Trial Matters Statistically
FRESCO-2 is a useful statistical teaching case because its registry results combine several fundamental clinical-trial methods in a single randomized oncology study: a time-to-event primary endpoint, ITT analysis, stratified log-rank testing, Cox hazard-ratio estimation, categorical-response analyses, and separate PK exposure-response modeling.
| Concept | How it appears in FRESCO-2 |
|---|---|
| Randomization | Participants were randomized to two parallel treatment groups. |
| Blinding | The registry classifies the trial as quadruple-masked. |
| ITT analysis | The primary OS analysis included all randomized participants. |
| Time-to-event endpoint | OS was the registered primary endpoint. |
| Stratified log-rank test | Used for the primary OS comparison and the secondary PFS comparison. |
| Hazard ratio | Used to quantify the relative OS and PFS treatment effects. |
| Confidence interval | 95% two-sided CIs were reported for the primary OS and secondary efficacy estimates. |
| Cochran-Mantel-Haenszel test | Used for ORR and DCR analyses. |
| Wald method | Used for confidence intervals around the adjusted differences for ORR and DCR. |
| Exposure-response analysis | Wald tests examined relationships between fruquintinib exposure and efficacy or safety parameters. |
The most important statistical lesson is that these methods answer different questions. The stratified Cox model describes a relative time-to-event effect; the Cochran-Mantel-Haenszel method addresses categorical outcomes while accounting for stratification; and the exposure-response analyses examine drug exposure relationships in PK-evaluable populations.
22. Statistical Methods Explained: Practical Reading Guide
What is the role of randomization?
Randomization creates the primary basis for comparing the treatment groups because treatment assignment is determined before post-randomization outcomes occur. The ITT analysis then preserves that randomized comparison. Randomization does not guarantee that every measured characteristic will be numerically identical between groups; its statistical purpose is to support unbiased comparison in expectation under the trial design.
Why does stratification appear in both randomization and analysis?
The registry states that the primary Cox analysis accounted for the randomization schedule stratification factors. Stratified analysis can therefore preserve information from the trial's randomized structure rather than ignoring the strata during the treatment comparison.
Why is a hazard ratio useful for OS?
OS contains information about both whether and when death occurs. A hazard ratio summarizes the relative event rate over the observed follow-up under the Cox model. This makes it possible to compare treatment groups without requiring every participant to have experienced the event by the time of analysis.
What is the difference between the OS and PFS hazard ratios?
The OS HR of 0.662 concerns death from any cause. The PFS HR of 0.321 concerns the first objective progression or death, whichever occurs first. Because the event definitions differ, the numerical HRs measure different outcomes and should not be compared as if one were simply a stronger version of the other.
Why is an adjusted difference used for ORR and DCR?
ORR and DCR are binary endpoints. The registry reports adjusted differences calculated using the Wald method from a Cochran-Mantel-Haenszel analysis that accounts for the randomization schedule stratification factors. This provides a stratified comparison rather than simply comparing two unadjusted percentages.
Why should the PK exposure-response analyses be kept separate?
The exposure-response analyses are based on pooled fruquintinib studies and PK-evaluable participants. Because this population is selected based on availability of PK parameter estimates, it is not equivalent to the randomized ITT population. The resulting Wald-test P-values therefore address exposure-response relationships rather than the primary randomized treatment effect.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: FRESCO-2, NCT04322539.
- Linked publication: PubMed record — PMID 40888992.
- Linked publication: PubMed record — PMID 40485075.
- Linked publication: PubMed record — PMID 40163688.
- Linked publication: PubMed record — PMID 39952149.
- Linked publication: PubMed record — PMID 37331369.
Continue with the statistical methods behind FRESCO-2
Explore the survival-analysis, categorical-data, clinical-trial, and statistical-inference methods that provide the framework for interpreting randomized trial results.
26. Record Summary
FRESCO-2 provides a clear example of a randomized phase 3 time-to-event analysis centered on overall survival. The ClinicalTrials.gov record reports an ITT primary analysis using a stratified log-rank test and a stratified Cox model, producing an OS hazard ratio of 0.662 with a two-sided 95% CI of 0.549–0.800 and P < .001. Secondary analyses extend the statistical picture to PFS, ORR, DCR, and exposure-response relationships.
The primary result is most appropriately understood as a combination of effect estimate, uncertainty, and hypothesis-test evidence. The HR describes the relative time-to-event effect; the confidence interval describes statistical precision; and the P-value addresses the prespecified statistical comparison. None of these quantities, by itself, provides the complete clinical picture.
The secondary endpoints also demonstrate why endpoint definitions matter. PFS is a time-to-event endpoint with progression or death as the event, while ORR and DCR are binary response measures analyzed with the Cochran-Mantel-Haenszel method. The exposure-response analyses are still further removed from the primary randomized comparison because they use pooled, PK-evaluable fruquintinib populations.