This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the registry-reported TASCO1 registry data.
1. Trial at a Glance
TASCO1 was a randomized, open-label, parallel phase 2 study in metastatic colorectal cancer. It compared trifluridine/tipiracil plus bevacizumab with capecitabine plus bevacizumab in patients with previously untreated disease who were non-eligible for intensive therapy.
| Feature | TASCO1 |
|---|---|
| Phase | Phase 2 |
| Condition | Metastatic Colorectal Cancer |
| Population | Patients with previously untreated colorectal cancer who were non-eligible for intensive therapy |
| Design | Randomized, parallel, unmasked |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 154 |
| Primary endpoint | Progression Free Survival (PFS) |
| Primary endpoint type | Time-to-event |
| ClinicalTrials.gov | NCT02743221 |
| Lead sponsor | Institut de Recherches Internationales Servier |
2. Clinical Question
The clinical question was whether, in previously untreated metastatic colorectal cancer patients who were non-eligible for intensive therapy, treatment with trifluridine/tipiracil plus bevacizumab produced a different time-to-progression-or-death profile from capecitabine plus bevacizumab.
Population
Patients with metastatic colorectal cancer who had not previously received treatment for the disease and were non-eligible for intensive therapy.
Intervention
Trifluridine/tipiracil plus bevacizumab.
Comparator
Capecitabine plus bevacizumab.
Primary question
How did progression-free survival compare between the two randomized treatment groups under the registered analysis?
3. Trial Design
Experimental treatment
- Trifluridine/tipiracil
- Bevacizumab
Comparator treatment
- Capecitabine
- Bevacizumab
The registry describes the allocation as randomized, the design model as parallel, and masking as none. Thus, this was an open-label randomized comparison rather than a blinded trial.
4. Trial Timeline
Study start
The registry lists April 29, 2016 as the study start date.
Primary completion
The registry lists January 15, 2018 as the primary completion date.
Registry status
The study is recorded as completed, with five statistical analyses posted across the reported outcome measures.
5. Primary Endpoint
| Endpoint | Registered definition / assessment | Analysis |
|---|---|---|
| Progression Free Survival (PFS) | Time from the date of randomisation until the date of investigator-assessed radiological disease progression or death due to any cause according to RECIST version 1.1. Assessments were scheduled at baseline and every 8 weeks, with a maximum follow-up duration of 17.9 months. | Cox proportional-hazards model adjusted for stratification factors |
6. Analysis Population and Stratification
The primary PFS analysis used the Full Analysis Set (FAS), defined in the registry as all randomized patients who had taken at least one dose of study treatment.
| Component | Registry information |
|---|---|
| Primary efficacy population | Full Analysis Set: all randomized patients who have taken at least one dose of study treatment |
| Randomization | Randomized |
| Primary analysis model | Cox proportional hazard model |
| Adjustment | Stratification factors |
| Stratification factors | RAS status and performance status ECOG |
| Effect measure | Hazard ratio |
This analysis population is important statistically. It is closely related to an intention-to-treat principle because patients remain associated with their randomized treatment assignment, but the registry-reported definition is not identical to a literal all-randomized ITT population: it additionally requires at least one dose of study treatment.
7. Primary Result: Progression-Free Survival
The primary endpoint was PFS, analyzed with a Cox proportional-hazards model adjusted for the stratification factors RAS status and performance status ECOG.
Hazard ratio for progression or death
95% CI: 0.48–1.06 · P = 0.09
Analysis population: Full Analysis Set
| Endpoint | Trifluridine/tipiracil + bevacizumab vs capecitabine + bevacizumab |
|---|---|
| Progression Free Survival | HR 0.71 (95% CI 0.48–1.06); P = 0.09 |
| Time frame | Baseline and every 8 weeks; maximum follow-up duration 17.9 months |
| Model | Cox proportional-hazards model |
| Adjustment | RAS status and performance status ECOG |
An HR of 0.71 means that the fitted Cox model estimated a lower instantaneous rate of progression or death in the trifluridine/tipiracil plus bevacizumab group relative to the capecitabine plus bevacizumab group. Expressed as a simple relative-hazard interpretation, 0.71 corresponds to an estimated 29% lower instantaneous hazard under the model.
The HR does not mean that 29% of patients avoided progression, that 29% of patients benefited, or that each patient experienced a 29% reduction in individual risk. It is a relative model-based measure of the event hazard.
The 95% CI of 0.48–1.06 describes uncertainty around the estimated HR. It includes 1, so the registry-reported interval is compatible with no difference in the modeled hazard as well as with a range of lower estimated hazards for the trifluridine/tipiracil combination.
The P-value of 0.09 is a measure of compatibility with the specified statistical comparison under its underlying assumptions; it is not a measure of the size or clinical importance of the treatment effect. A P-value cannot tell us whether an HR of 0.71 is large or small.
Finally, the registry describes the study as non-comparative and does not state a superiority or non-inferiority hypothesis. Therefore, the PFS result should not be reframed as a formal non-inferiority conclusion. Interpretation of the Cox HR also depends on the model's proportional-hazards assumption.
8. Secondary Endpoint Results
Overall Survival
Overall survival was evaluated from baseline up to death or study cut-off, with a maximum follow-up duration of 19.9 months. The analysis population was the FAS, and the registered method was a Cox proportional-hazards model.
Hazard ratio for overall survival
95% CI: 0.32–0.98 · P = 0.04
An HR of 0.56 indicates that the fitted model estimated a lower instantaneous rate of death in the trifluridine/tipiracil plus bevacizumab group relative to the capecitabine plus bevacizumab group. As a direct mathematical interpretation of the estimate, 0.56 corresponds to approximately a 44% lower estimated hazard.
This does not mean that 44% of patients lived longer, that survival probability was 44% higher, or that an individual patient's probability of death was reduced by exactly 44%. Hazard ratios describe relative event rates within a time-to-event model.
The 95% CI of 0.32–0.98 indicates substantial uncertainty around the point estimate. The upper endpoint remains below 1, while the interval is nevertheless fairly broad relative to the point estimate.
The P-value of 0.04 should not be interpreted as a 4% probability that the treatment effect is real, nor as a measure of effect magnitude. It describes the statistical evidence under the analysis framework, not the clinical size of the effect.
Because the registry does not state a conventional superiority hypothesis and identifies the study as non-comparative, the result should be described as a reported Cox-model estimate rather than transformed into a stronger causal or confirmatory claim than the registry supports.
Duration of Response
Duration of Response (DR) was a secondary time-to-event endpoint assessed at baseline and every 8 weeks, with a maximum follow-up duration of 16.6 months. The analysis population was the Tumour Response population: patients with measurable disease at baseline and at least one tumour evaluation while on treatment.
Hazard ratio for duration of response
95% CI: 0.49–2.74
The reported HR of 1.16 is a relative time-to-event estimate for duration of response comparing trifluridine/tipiracil plus bevacizumab with capecitabine plus bevacizumab. Unlike an HR below 1, an HR above 1 means the modeled event hazard was estimated to be higher in the first group under the direction used for the reported comparison.
The wide 95% CI of 0.49–2.74 spans 1 by a substantial amount. Consequently, the estimate is imprecise and does not by itself establish a difference in duration of response.
No formal analysis method is reported for DR in the registry statistical-analysis record. The endpoint is nevertheless a time-to-event measure, for which Kaplan-Meier estimation and a time-to-event model would ordinarily be considered depending on the prespecified analysis plan. This page does not attribute an unreported method to the TASCO1 registry.
Overall Response Rate
Overall Response Rate (ORR) was assessed at baseline and every 8 weeks, with a maximum follow-up duration of 17.9 months. The analysis population was the FAS, and the registry reports a Fisher exact test.
The ClinicalTrials.gov record reports the P-value but does not provide arm-specific response counts or response percentages. This page therefore does not construct an ORR estimate from unavailable data.
Fisher's exact test is appropriate for comparing a binary outcome between two groups when an exact calculation of the contingency-table probability is desired. The P-value of 0.73 is not a measure of response magnitude and cannot be converted into an ORR difference without the underlying response counts.
Disease Control Rate
Disease Control Rate (DCR) was assessed at baseline and every 8 weeks, with a maximum follow-up duration of 17.9 months. The analysis population was the FAS, and the registry reports a Fisher exact test.
The ClinicalTrials.gov record reports the P-value but does not provide arm-specific disease-control counts or percentages. No effect estimate is therefore added.
As with ORR, the Fisher exact P-value addresses the statistical comparison of the binary outcome under the reported test. It does not state the absolute disease-control rate in either arm or quantify how large any difference was.
9. Statistical Methodology
Cox proportional-hazards model
The primary PFS and secondary OS analyses used a Cox proportional-hazards model. The PFS model was explicitly adjusted for the stratification factors RAS status and performance status ECOG. The effect measure was the hazard ratio.
The hazard ratio associated with a treatment indicator can be expressed as exp(β). It compares the modeled instantaneous event hazards while accounting for the covariates included in the model.
The central modeling assumption is proportional hazards: the relative hazard represented by the HR is assumed to be reasonably stable over time. A single HR can be difficult to summarize if the relative hazards change substantially during follow-up. The ClinicalTrials.gov record does not provide diagnostics for this assumption, so no conclusion about its adequacy is made here.
Covariate adjustment
The PFS analysis adjusted for RAS status and performance status ECOG, which were the identified stratification factors. Adjustment can improve the statistical efficiency of a randomized comparison when important prognostic factors are included appropriately.
Adjustment does not mean that randomization failed. Rather, the model incorporates prespecified information about clinically relevant characteristics into estimation of the treatment-associated hazard ratio.
Fisher exact test
ORR and DCR were binary endpoints, and the registry reports Fisher exact tests for both. The test works directly with the observed two-by-two treatment-by-outcome table and calculates an exact probability under the null framework.
The test produces a P-value for the comparison, but the P-value does not itself provide the response rates, risk difference, risk ratio, or odds ratio.
Intention-to-treat principle and the FAS
The analysis record includes intention-to-treat analysis among the concepts associated with the PFS analysis. However, the actual stated population is the FAS: all randomized patients who had taken at least one dose of study treatment.
This distinction is worth preserving. A strict ITT analysis generally retains every randomized participant according to randomized assignment, whereas the registry-reported FAS definition introduces a post-randomization treatment-exposure condition. The registry wording should therefore be retained rather than silently relabeling the FAS as pure ITT.
Stratified analysis
Stratification is reflected in the primary PFS Cox model through adjustment for RAS status and performance status ECOG. This allows the estimated treatment comparison to account for these prespecified factors rather than ignoring them in the primary model.
10. Statistical Methods Explained
Why was a Cox model used for PFS?
PFS is a time-to-event endpoint: participants can experience progression or death at different times, while others may remain event-free at their last assessment. A Cox model uses the timing information and accommodates censoring rather than reducing the endpoint to a simple yes/no outcome.
What does an HR of 0.71 mean?
An HR of 0.71 means the fitted model estimated the instantaneous progression-or-death hazard to be 0.71 times that of the comparator, under the model and follow-up represented by the analysis. It does not mean that 71% of patients remained progression-free or that every patient's individual risk was reduced by 29%.
Why adjust for RAS status and ECOG performance status?
The primary PFS analysis was adjusted for these two stratification factors. Covariate adjustment can account for prognostically important baseline information and can make the estimated treatment comparison more efficient. It does not replace randomization; it uses additional information within the randomized comparison.
Why is the PFS confidence interval important?
The HR of 0.71 is only one estimate. The 95% CI of 0.48–1.06 communicates the uncertainty around that estimate. Because the interval crosses 1, the reported data do not isolate a single direction of the treatment hazard effect with high precision.
Why does a P-value not measure treatment effect size?
A P-value summarizes evidence against a specified null framework under the assumptions of the test. It depends on the effect, variability, and amount of information. An HR, an absolute difference, or another effect measure is needed to describe magnitude. Thus P = 0.09 and P = 0.04 should not be read as "small" and "large" effects, respectively.
What does Fisher's exact test contribute for ORR and DCR?
Fisher's exact test evaluates the association between treatment group and a binary outcome using the observed contingency table. It supplies a P-value, but without the underlying counts it does not tell the reader the actual response or disease-control rates in the two groups.
Why is the non-comparative designation important?
The ClinicalTrials.gov record states that the study was a non-comparative study in the field where a non-inferiority or equivalence margin might otherwise be recorded. No non-inferiority margin is reported and the hypothesis type is listed as other/not stated. Therefore, the reported HRs and P-values should not be converted into a non-inferiority claim or a conventional superiority conclusion that is not specified in the registry data.
11. Primary and Secondary Results at a Glance
| Endpoint | Role | Analysis | Estimate / P-value | 95% CI |
|---|---|---|---|---|
| Progression Free Survival | Primary | Cox proportional-hazards model | HR 0.71; P = 0.09 | 0.48–1.06 |
| Duration of Response | Secondary | Method not reported | HR 1.16 | 0.49–2.74 |
| Overall Survival | Secondary | Cox proportional-hazards model | HR 0.56; P = 0.04 | 0.32–0.98 |
| Overall Response Rate | Secondary | Fisher exact test | P = 0.73 | Not reported |
| Disease Control Rate | Secondary | Fisher exact test | P = 0.22 | Not reported |
The table illustrates why statistical interpretation should not be reduced to a single P-value. TASCO1 reports different endpoint types, different analysis methods, different analysis populations, and different levels of available effect-measure information.
12. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm. These figures are presented as affected participants / participants at risk.
| Safety measure | Trifluridine/tipiracil + bevacizumab | Capecitabine + bevacizumab |
|---|---|---|
| Serious adverse events | 42 / 77 | 44 / 76 |
The figures describe the number of participants affected and the corresponding at-risk group reported in the registry. They are not accompanied in the ClinicalTrials.gov record by a formal statistical comparison, confidence interval, or P-value, so none is added here.
13. Understanding the Primary PFS Estimate
The primary PFS HR of 0.71 is a relative measure. Within the Cox model, it represents the estimated instantaneous hazard of progression or death in the trifluridine/tipiracil plus bevacizumab group relative to the capecitabine plus bevacizumab group.
The ClinicalTrials.gov record does not report median PFS, landmark PFS probabilities, or other absolute PFS estimates. Consequently, this page does not attempt to translate the HR into an absolute number of additional months or a percentage-point difference.
The 95% CI of 0.48–1.06 is relatively broad around the HR point estimate. The lower endpoint represents a considerably lower estimated hazard, while the upper endpoint is slightly above 1. The interval therefore communicates meaningful uncertainty about the magnitude and even the direction of the modeled effect.
The P-value of 0.09 should be interpreted within the registry's stated framework. Because the hypothesis type is listed as other/not stated and the study is described as non-comparative, it should not be treated as evidence for or against an unstated superiority or non-inferiority hypothesis.
14. Why the Different Endpoints Require Different Statistics
| Endpoint type | TASCO1 endpoint | Reported method | Why the method differs |
|---|---|---|---|
| Time-to-event | PFS | Cox proportional-hazards model | Uses event timing and accommodates censoring. |
| Time-to-event | OS | Cox proportional-hazards model | Models time until death while accounting for censoring. |
| Time-to-event | Duration of Response | Not reported | Requires time-to-event methodology when analyzed formally. |
| Binary | ORR | Fisher exact test | Compares binary response classifications between groups. |
| Binary | DCR | Fisher exact test | Compares binary disease-control classifications between groups. |
This distinction is fundamental. A time-to-event endpoint contains information about when an event occurred and whether observation ended before an event. A binary endpoint collapses the observation into a classification such as response versus no response. The statistical method should reflect that structure.
15. Non-Inferiority, Equivalence, and the TASCO1 Analysis Record
Non-inferiority trials require a prespecified margin defining how much loss of efficacy would still be considered acceptable relative to the comparator. Equivalence trials similarly require prespecified bounds around the treatment difference. Neither concept can be established simply by observing that a confidence interval crosses or does not cross a numerical threshold chosen after the fact.
For TASCO1, the ClinicalTrials.gov record explicitly states "This was a non-comparative study." The hypothesis type is listed as Other / not stated, and no non-inferiority or equivalence margin is provided.
16. Multiplicity and Multiple Endpoints
TASCO1 has one registered primary endpoint, PFS, and four additional reported secondary analyses: Duration of Response, Overall Survival, Overall Response Rate, and Disease Control Rate.
| Endpoint | Role | Reported statistical information |
|---|---|---|
| Progression Free Survival | Primary | HR 0.71; 95% CI 0.48–1.06; P = 0.09 |
| Duration of Response | Secondary | HR 1.16; 95% CI 0.49–2.74 |
| Overall Survival | Secondary | HR 0.56; 95% CI 0.32–0.98; P = 0.04 |
| Overall Response Rate | Secondary | Fisher exact P = 0.73 |
| Disease Control Rate | Secondary | Fisher exact P = 0.22 |
The presence of a P-value of 0.04 for a secondary endpoint does not, by itself, establish that the secondary endpoint was tested under a multiplicity-controlled confirmatory framework. The ClinicalTrials.gov record does not specify an alpha-allocation strategy, hierarchical testing sequence, or multiplicity adjustment, so none is assumed.
17. Interim Analysis and Missing Data
The registry-reported TASCO1 data does not report an interim-analysis procedure, alpha-spending method, stopping boundary, missing-data imputation method, or detailed censoring rules beyond the time-to-event endpoint descriptions.
Interim analysis
No interim-analysis method is reported in the ClinicalTrials.gov record. No group-sequential boundary or alpha-spending rule is added.
Missing data
No imputation strategy is reported in the ClinicalTrials.gov record. No specific missing-data assumption is therefore attributed to the trial.
This is an important methodological boundary. Statistical analysis pages should distinguish between methods that are actually documented and methods that would merely be plausible for a trial of this type.
18. Randomization and Causal Interpretation
Randomization is a central feature of TASCO1's design. By assigning participants to treatment groups randomly, the design aims to make treatment assignment independent of measured and unmeasured baseline characteristics in expectation.
That design property supports a treatment comparison, but it does not eliminate every source of uncertainty. The actual analysis population, treatment exposure, censoring, follow-up duration, model assumptions, and subsequent clinical events all influence the observed estimate.
What randomization provides
A principled basis for comparing treatment groups without deliberately assigning treatment according to prognosis.
What randomization does not provide
It does not make an HR exact, remove sampling uncertainty, or guarantee that every baseline characteristic is numerically identical between groups.
19. What the Hazard Ratio Does — and Does Not — Mean
The PFS HR of 0.71 means that, under the fitted Cox model, the estimated instantaneous hazard of progression or death was 0.71 times that of the comparator group. This can be described as approximately a 29% lower estimated hazard, but not as a 29% reduction in every patient's individual probability of progression.
The OS HR of 0.56 corresponds to an estimated instantaneous death hazard approximately 44% lower in the trifluridine/tipiracil plus bevacizumab group under the fitted model. It does not mean that survival probability increased by 44% or that 44% of participants benefited.
The PFS CI of 0.48–1.06 and OS CI of 0.32–0.98 describe uncertainty around their respective HR estimates. They are not ranges containing the effects experienced by individual patients.
The PFS P-value of 0.09 and OS P-value of 0.04 provide statistical-test information. They do not measure the size, importance, or probability of a treatment effect.
20. Limitations
- Non-comparative designation: the registry analysis record describes TASCO1 as a non-comparative study and does not state a conventional superiority or non-inferiority hypothesis.
- No non-inferiority margin: no non-inferiority or equivalence margin is reported, so no margin-based conclusion can be reconstructed.
- Primary PFS precision: the 95% CI of 0.48–1.06 spans 1, indicating uncertainty around the primary HR estimate.
- Secondary endpoint interpretation: the OS result has a reported P-value of 0.04, but the ClinicalTrials.gov record does not document multiplicity control or an endpoint hierarchy that would justify treating it as an independently confirmatory result.
- Incomplete response detail: ORR and DCR P-values are reported, but arm-specific response counts or percentages are not reported.
- Duration-of-response method: a hazard ratio and confidence interval are reported for DR, but the specific statistical method is not identified in the ClinicalTrials.gov record.
- Analysis population: the FAS requires at least one dose of study treatment, so it is not identical to a literal all-randomized ITT population.
- Cox assumptions: interpretation of the PFS and OS HRs depends on the proportional-hazards model being an appropriate representation of the data.
- Missing-data methodology: the ClinicalTrials.gov record does not specify an imputation strategy or detailed missing-data framework.
- Limited safety detail: serious adverse-event counts by arm are available, but detailed event-level safety results and formal safety comparisons are not included in the ClinicalTrials.gov record.
21. Why This Trial Matters Statistically
TASCO1 is a useful teaching example because it brings together several recurring issues in clinical-trial statistics without relying on a single endpoint or a single statistical test.
| Concept | How it appears in TASCO1 |
|---|---|
| Randomization | The study used randomized allocation in a parallel design. |
| Time-to-event analysis | PFS, OS, and DR were analyzed or reported as time-to-event outcomes. |
| Hazard ratio | PFS, OS, and DR were reported using HR effect measures. |
| Cox model | PFS and OS used Cox proportional-hazards models. |
| Covariate adjustment | The PFS Cox model adjusted for RAS status and ECOG performance status. |
| Fisher exact test | ORR and DCR were analyzed using Fisher exact tests. |
| Confidence intervals | 95% two-sided CIs were reported for the PFS, OS, and DR HRs. |
| FAS analysis | The primary PFS analysis used all randomized patients who took at least one dose. |
| Stratified analysis | RAS status and ECOG performance status were used as stratification factors in the PFS model. |
| Multiplicity | One primary endpoint was accompanied by four reported secondary analyses, without a registry-reported multiplicity strategy. |
| Non-inferiority logic | No non-inferiority margin was reported; the study was described as non-comparative. |
22. A Practical Reading of the TASCO1 Statistical Record
A disciplined reading starts with the endpoint rather than the P-value. PFS is a time-to-event outcome, so the appropriate effect measure is a hazard ratio from a time-to-event model. The reported estimate is 0.71, with a two-sided 95% CI of 0.48–1.06.
The next question is precision. The interval is wide enough to encompass both a meaningfully lower estimated hazard and a value slightly above 1. That tells the reader more about the uncertainty of the result than the P-value alone.
The third question is what hypothesis was actually specified. The registry identifies the hypothesis type as other/not stated and explicitly records that the study was non-comparative. Without a stated non-inferiority margin or superiority hypothesis, it would be inappropriate to retrofit a conventional confirmatory interpretation onto the result.
The secondary endpoints then require separate interpretation. OS has an HR of 0.56 with a 95% CI of 0.32–0.98 and P = 0.04. DR has an HR of 1.16 with a 95% CI of 0.49–2.74. ORR and DCR have Fisher exact P-values of 0.73 and 0.22, respectively. These numbers should not be collapsed into a single overall statistical judgment because they describe different outcomes using different analytical structures.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Statistical Calculators
25. Sources
- ClinicalTrials.gov: TASCO1 — NCT02743221. Official registry record and source for the trial data presented on this page.
- Linked publication: PubMed record — PMID 32497736.
Continue through the Clinical Biostats statistical library
Explore the statistical concepts behind randomized trials, survival analysis, categorical endpoints, confidence intervals, and clinical-trial interpretation.
26. Record Summary
TASCO1 was a randomized phase 2 study enrolling 154 participants and comparing trifluridine/tipiracil plus bevacizumab with capecitabine plus bevacizumab in previously untreated metastatic colorectal cancer patients who were non-eligible for intensive therapy. Its primary endpoint was PFS, assessed at baseline and every 8 weeks, with a maximum follow-up duration of 17.9 months. The primary analysis used a Cox proportional-hazards model adjusted for RAS status and ECOG performance status.
The reported primary PFS estimate was HR 0.71 with a two-sided 95% CI of 0.48–1.06 and P = 0.09. Secondary analyses reported an OS HR of 0.56 (95% CI 0.32–0.98; P = 0.04), a DR HR of 1.16 (95% CI 0.49–2.74), an ORR Fisher exact P-value of 0.73, and a DCR Fisher exact P-value of 0.22.
The statistical interpretation depends on more than these estimates. The study is recorded as non-comparative, the hypothesis type is not stated as a conventional superiority or non-inferiority hypothesis, and no non-inferiority margin or multiplicity strategy is reported in the ClinicalTrials.gov record. The analysis population for PFS is the FAS rather than a literal all-randomized ITT population. These design and analysis details define the limits of what can be inferred from the reported results.