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
NO16968 was a randomized, parallel-group, open-label phase 3 trial in colorectal cancer. The registry reports 1886 enrolled participants and two treatment arms, with disease-free survival as the primary time-to-event endpoint formally analyzed using a log-rank test and hazard ratio.
| Feature | NO16968 |
|---|---|
| Trial name | NO16968 |
| ClinicalTrials.gov identifier | NCT00069121 |
| Brief title | A Study of Xeloda (Capecitabine) Plus Oxaliplatin in Patients With Colon Cancer |
| Therapeutic area | Oncology |
| Condition | Colorectal Cancer |
| Phase | Phase 3 |
| Status | Completed |
| Start | April 18, 2003 |
| Primary completion | April 21, 2011 |
| Enrollment | 1886 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Lead sponsor | Hoffmann-La Roche |
| Sponsor type | Industry |
2. Clinical Question
The registered trial compares XELOX with 5-FU/LV in patients with colorectal cancer. The principal statistical question is whether the time from randomization to a disease-free survival event differs between the randomized treatment groups.
Population
Participants enrolled in the NO16968 phase 3 study with colorectal cancer.
Intervention
XELOX, represented in the registry by capecitabine plus oxaliplatin.
Comparator
5-FU/LV, represented by 5-fluorouracil plus leucovorin (LV).
Primary question
Does disease-free survival differ between XELOX and 5-FU/LV after randomization?
3. Trial Design
The registry describes NO16968 as a randomized, parallel-group, unmasked phase 3 treatment trial. Four interventions are listed: capecitabine, oxaliplatin, leucovorin (LV), and 5-fluorouracil (5-FU).
Capecitabine + oxaliplatin
- Capecitabine
- Oxaliplatin
- Registry treatment group used as the numerator/reference direction for reported hazard ratios
5-fluorouracil + leucovorin
- 5-Fluorouracil (5-FU)
- Leucovorin (LV)
- Comparator treatment group for the reported efficacy analyses
4. Analysis Population
The posted statistical analyses use the Intent-to-Treat (ITT) Population. The registry defines this as all randomized participants who gave written informed consent and states that participants were analyzed according to the treatment arm to which they were randomized.
| Analysis population | Registry description / role |
|---|---|
| Intent-to-Treat | All randomized participants who gave written informed consent; participants were analyzed according to randomized treatment arm. |
| Safety population | The ClinicalTrials.gov record reports serious adverse events by the listed treatment categories; a separate formal safety-population definition is not reported. |
The ITT framework is particularly important for interpreting the efficacy results because randomization establishes the treatment groups before subsequent follow-up, events, censoring, or treatment changes occur. An ITT comparison therefore preserves the original randomized treatment contrast rather than redefining groups according to treatment actually received.
5. Endpoints
| Endpoint | Registry time frame | Endpoint type |
|---|---|---|
| Disease-Free Survival (DFS) [Number of Events] | Time from randomization date to date of first DFS event/date last known to be event-free. Median observation time for DFS was 74 months (range: 0-95 months). | Time-to-event |
| Disease-Free Survival (DFS) [Time to Event] | Time from randomization date to date of first DFS event/date last known to be event-free. Median observation time for DFS was 74 months (range: 0-95 months). | Time-to-event |
| Relapse-Free Survival (RFS) [Time to Event] | Time from randomization date to date of first RFS event/date last known to be event-free. Median observation time for RFS was 74 months (range: 0-95 months). | Time-to-event |
| Overall Survival [Time to Event] | Time from randomization date to date of death/date last known to be alive. Median observation time was 83 months (range: 0-95 months). | Time-to-event |
What counted as a DFS event?
The registry states that determination of an event was based on tumor assessments and survival follow-up assessments. A DFS event was defined as any recurrence of the original colon cancer, appearance of a new colon or rectal cancer, or death due to any cause, whichever was earliest. New colon or rectal cancer was to be proven by cytology or histology when possible. Participants without such an event at the time of data assessment contributed follow-up through the date they were last known to be event-free.
The distinction between the two registered DFS measures is important. One is explicitly labeled [Number of Events], while the other is [Time to Event]. The posted formal statistical comparison is for the time-to-event DFS endpoint.
6. Statistical Methodology
Log-rank test
The registry identifies the log-rank test as the statistical method for the posted time-to-event analyses. The log-rank test compares the survival experience of two groups across the observed follow-up period while accounting for the timing of events and right censoring.
For a time-to-event endpoint such as DFS, the log-rank test uses the ordering and timing of events rather than reducing every participant to a simple event/no-event proportion at a fixed time.
Hazard ratio
The registry reports the hazard ratio as the effect measure for DFS, RFS, and overall survival. For each analysis, the HR is explicitly calculated as XELOX versus 5-FU/LV.
An HR is a relative time-to-event measure. It is not the same as a relative risk, an absolute risk difference, a probability of cure, or a statement that every participant experiences the same proportional change in risk.
Intention-to-treat analysis
All three posted efficacy analyses use the ITT population. An ITT analysis maintains the randomized treatment assignment when comparing outcomes. This is useful because the treatment groups were created by randomization rather than by subsequent patient characteristics or observed treatment response.
Censoring
The registered endpoint definitions explicitly include dates at which participants were last known to be event-free or alive. This is the fundamental structure of right-censored time-to-event data: a participant who has not yet experienced the endpoint still contributes information up to the point at which follow-up ends or the participant is otherwise censored.
Two-sided confidence intervals
The primary DFS hazard ratio is accompanied by a 95% two-sided confidence interval. Confidence intervals provide information about the statistical precision of an estimated treatment effect; they are not ranges containing a specified percentage of individual patients' true outcomes.
7. Primary Results
The registry contains formal statistical analysis for the primary Disease-Free Survival [Time to Event] endpoint. A second registered primary measure, DFS [Number of Events], is marked as having results posted, but the ClinicalTrials.gov record does not provide a separate formal statistical comparison for the event count.
Primary Endpoint: Disease-Free Survival [Time to Event]
Hazard ratio for disease-free survival
95% CI: 0.69–0.93 · P = 0.0038
Two-sided 95% confidence interval · Log-rank test · ITT population
| Primary DFS analysis | Reported result |
|---|---|
| Groups compared | 5-FU/LV vs XELOX |
| Hazard ratio direction | XELOX versus 5-FU/LV |
| Hazard ratio | 0.80 |
| 95% CI | 0.69–0.93 |
| Confidence interval | Two-sided |
| P-value | 0.0038 |
| Method | Log-rank test |
| Hypothesis type | Superiority |
| Analysis population | Intent-to-Treat |
The reported HR of 0.80 means that, in the time-to-event comparison specified by the registry, the estimated instantaneous rate of a DFS event for XELOX was approximately 20% lower than for 5-FU/LV. That is a relative hazard interpretation; it does not mean that exactly 20% fewer participants experienced an event, nor that each individual participant had a 20% reduction in risk.
The 95% CI of 0.69–0.93 describes uncertainty around the estimated hazard ratio under the statistical framework used for the analysis. Its width gives information about precision. The interval does not describe the range of individual treatment effects across patients.
The P-value of 0.0038 addresses evidence against the null hypothesis represented by the statistical test. It is not a measure of how large the treatment effect is. Effect magnitude is conveyed by the HR and its confidence interval, while the p-value reflects the compatibility of the observed data with the tested null under the model and assumptions.
The registry identifies a log-rank analysis and a hazard ratio but does not supply additional model diagnostics in the ClinicalTrials.gov record. Accordingly, the HR should be interpreted as the reported time-to-event summary rather than as proof that the hazard ratio was constant at every point in follow-up.
Primary Endpoint: Disease-Free Survival [Number of Events]
For a time-to-event endpoint, simply comparing the final number of events can discard information about when those events occurred and how long participants were followed. A survival-analysis framework is therefore generally more informative than a simple event-count comparison when follow-up times differ or censoring is present.
8. Secondary Results
Relapse-Free Survival [Time to Event]
Hazard ratio for relapse-free survival
95% CI: 0.67–0.91 · P = 0.0015
Two-sided 95% confidence interval · Log-rank test · ITT population
| RFS analysis | Reported result |
|---|---|
| Groups compared | 5-FU/LV vs XELOX |
| Hazard ratio direction | XELOX versus 5-FU/LV |
| Hazard ratio | 0.78 |
| 95% CI | 0.67–0.91 |
| P-value | 0.0015 |
| Method | Log-rank test |
| Hypothesis type | Other / not stated |
| Registry analysis note | Descriptive analysis only. |
The HR of 0.78 corresponds to an estimated instantaneous RFS event rate approximately 22% lower with XELOX than with 5-FU/LV under the reported comparison. The registry explicitly labels this analysis as descriptive only, so the numerical p-value should not be treated as though this secondary analysis were assigned the same confirmatory role as the primary superiority analysis.
Overall Survival [Time to Event]
Hazard ratio for overall survival
95% CI: 0.70–0.99 · P = 0.0367
Two-sided 95% confidence interval · Log-rank test · ITT population
| Overall survival analysis | Reported result |
|---|---|
| Groups compared | 5-FU/LV vs XELOX |
| Hazard ratio direction | XELOX versus 5-FU/LV |
| Hazard ratio | 0.83 |
| 95% CI | 0.70–0.99 |
| P-value | 0.0367 |
| Method | Log-rank test |
| Hypothesis type | Other / not stated |
| Registry analysis note | Descriptive analysis only. |
| Registered time frame | Time from randomization date to date of death/date last known to be alive. Median observation time was 83 months (range: 0-95 months). |
The OS HR of 0.83 corresponds to an estimated instantaneous rate of death approximately 17% lower with XELOX than with 5-FU/LV under the reported time-to-event comparison. The 95% CI extends from 0.70 to 0.99, so the reported estimate is more precise around a relatively modest relative hazard difference than the larger effects that would be represented by a substantially smaller HR.
The three posted hazard ratios all have the same direction because the registry defines each as XELOX versus 5-FU/LV: 0.80 for DFS, 0.78 for RFS, and 0.83 for OS. They are not interchangeable endpoints. DFS, RFS, and OS use different event definitions, so each HR answers a different time-to-event question.
The registry labels the RFS and OS analyses as descriptive analyses. This is an important distinction from the formally analyzed primary DFS endpoint: a nominal p-value in a secondary descriptive analysis does not by itself establish that the endpoint was part of the same confirmatory hypothesis-testing hierarchy.
9. Safety
The ClinicalTrials.gov record reports serious adverse events by the listed treatment categories. The safety figures are presented as affected participants / participants at risk.
| Listed treatment category | Serious adverse events | Affected / at risk |
|---|---|---|
| 5-FU/LV MAYO CLINIC | Serious adverse events | 134 / 657 |
| 5-FU/LV ROSWELL PARK | Serious adverse events | 88 / 269 |
| XELOX | Serious adverse events | 208 / 938 |
The ClinicalTrials.gov record lists two separate 5-FU/LV categories alongside XELOX, even though the overall ClinicalTrials.gov record identifies two arms. Because the ClinicalTrials.gov record does not explain the relationship between those safety categories and the two-arm design, they should be reported as registered rather than collapsed into a newly calculated 5-FU/LV total.
10. Statistical Methods Explained
Why was a log-rank test used?
The primary endpoint is a time-to-event outcome. Participants can experience a DFS event at different times, while others may remain event-free at the end of their available follow-up. The log-rank test is designed to compare the survival experience of two groups while retaining information about the timing of events and censoring.
What does an HR of 0.80 mean?
Because the registry defines the HR as XELOX versus 5-FU/LV, an HR of 0.80 represents a lower estimated instantaneous DFS event rate in the XELOX group. Expressed as a simple relative interpretation, \(1-0.80=0.20\), or approximately a 20% lower estimated hazard. It does not mean that 20% of patients were protected from recurrence or death.
Why does the confidence interval matter?
The DFS 95% CI is 0.69–0.93. The point estimate of 0.80 is therefore only one estimate within an uncertainty interval. A confidence interval communicates the precision of the estimated treatment effect more directly than a p-value alone. It also shows that the registry's estimated effect is not a single exact quantity.
Why does the p-value not measure effect size?
The p-value of 0.0038 for DFS quantifies evidence relative to the statistical null hypothesis under the specified test. It does not say that the treatment effect is 0.0038 in magnitude, nor does a smaller p-value automatically represent a larger clinical effect. The HR and its confidence interval are the appropriate reported measures for the magnitude and precision of this time-to-event comparison.
Why is ITT analysis important?
The posted analyses use all randomized participants who gave written informed consent and retain their randomized treatment assignment. This protects the comparison created by randomization. If participants were instead reclassified after randomization based on treatment received or later clinical events, the groups could become systematically different in ways that randomization was intended to control.
What does censoring mean for DFS and OS?
A participant who has not experienced the event by the last date on which the participant is known to be event-free or alive has not necessarily demonstrated that the event will never occur. Instead, the participant's follow-up is censored at that point. Survival-analysis methods use the information available up to censoring rather than treating such participants as if they had completed the entire observation period without an event.
Why should the secondary p-values be interpreted differently?
The registry describes the RFS and OS analyses as descriptive analysis only, while the primary DFS analysis is explicitly associated with a superiority hypothesis. This distinction matters because statistical significance is not interpreted in isolation from the role assigned to an endpoint in the trial's analysis plan. The ClinicalTrials.gov record does not provide a multiplicity strategy for these secondary analyses, so a descriptive p-value should not be elevated into an unreported confirmatory claim.
11. Understanding the Primary DFS Result
The primary DFS HR of 0.80 indicates an estimated 20% lower instantaneous event rate for XELOX versus 5-FU/LV, based on the direction specified by the registry. The estimate summarizes relative event rates over follow-up; it does not provide an absolute percentage of patients who benefited.
The 0.69–0.93 95% confidence interval shows the uncertainty associated with the estimated HR. It is narrower than an interval that would arise from a very imprecise estimate, but it still contains a range of plausible values rather than a single known treatment effect.
The reported P = 0.0038 is associated with the superiority hypothesis and the log-rank analysis. It provides evidence against the corresponding null hypothesis under the test framework. It should not be converted into a probability that the treatment works, nor into an estimate of clinical importance.
An HR is a model-based summary of relative instantaneous event rates. Without additional information about the shape of the underlying event curves, a single HR should not be treated as though it describes the relative risk at every individual time point.
12. DFS Number of Events vs DFS Time to Event
NO16968 illustrates an important distinction in clinical-trial reporting: the same clinical concept can be represented both as an event count and as a time-to-event endpoint.
| Representation | What it captures | Statistical information reported |
|---|---|---|
| DFS [Number of Events] | Whether a defined DFS event occurred, summarized as a number of events | Results posted, but no formal comparison reported in the statistical-analyses data |
| DFS [Time to Event] | Time from randomization to the first DFS event or last known event-free date | Formal log-rank analysis, HR 0.80, 95% CI 0.69–0.93, P = 0.0038 |
The time-to-event formulation is statistically richer because two trials or treatment groups could have the same eventual number of events but substantially different timing of those events. The survival-analysis approach uses both event occurrence and event timing while accommodating censoring.
13. Relationship Among DFS, RFS, and Overall Survival
The registry provides three related but distinct time-to-event outcomes. Their definitions determine what each analysis measures.
Disease-Free Survival
Begins at randomization and ends at the first qualifying DFS event or the date last known to be event-free. DFS includes recurrence, new colon or rectal cancer, or death from any cause.
Relapse-Free Survival
Begins at randomization and ends at the first RFS event or the date last known to be event-free. The registry definition identifies the time frame but does not provide a complete event-definition text.
Overall Survival
Begins at randomization and ends at death or the date last known to be alive. It therefore has a different event definition from DFS and RFS.
Why the HRs differ
DFS, RFS, and OS are not interchangeable. Their HRs can differ because they count different events and therefore represent different statistical questions.
14. Comparing Relative Effects
| Endpoint | HR, XELOX vs 5-FU/LV | 95% CI | P-value | Registry role |
|---|---|---|---|---|
| Disease-Free Survival | 0.80 | 0.69–0.93 | 0.0038 | Primary; superiority |
| Relapse-Free Survival | 0.78 | 0.67–0.91 | 0.0015 | Secondary; descriptive analysis only |
| Overall Survival | 0.83 | 0.70–0.99 | 0.0367 | Secondary; descriptive analysis only |
All three estimates are below 1, so all three point estimates favor XELOX in the direction of the registered hazard-ratio calculation. The estimates should nevertheless be interpreted separately because the endpoints have different event definitions and different roles in the statistical framework.
The closeness of the HRs should not be interpreted as evidence that the endpoints are statistically identical. Nor should the p-values be ranked as though the smallest p-value represented the most important endpoint. The primary-versus-secondary designation and the registered analysis notes are essential parts of the interpretation.
15. What the Registry Does and Does Not Establish
| Question | What the ClinicalTrials.gov record supports |
|---|---|
| Was the trial randomized? | Yes. Allocation is registered as randomized. |
| Was the design parallel? | Yes. The design model is registered as parallel. |
| Was the trial masked? | No. Masking is registered as none. |
| Was DFS formally analyzed? | Yes. The DFS time-to-event endpoint has a posted log-rank analysis and HR. |
| Was DFS an event-count endpoint? | Yes. A primary DFS number-of-events endpoint is registered and results are posted, but no separate formal comparison is reported in the statistical-analyses data. |
| Were RFS and OS analyzed? | Yes. Both have posted log-rank analyses with hazard ratios and 95% confidence intervals. |
| Was the RFS analysis confirmatory? | The ClinicalTrials.gov record labels it descriptive analysis only. |
| Was the OS analysis confirmatory? | The ClinicalTrials.gov record labels it descriptive analysis only. |
| Was a non-inferiority margin reported? | No non-inferiority margin is reported. The DFS hypothesis type is superiority. |
| Was crossover reported? | No crossover information is reported in the ClinicalTrials.gov record. |
| Was an interim analysis reported? | No interim-analysis information is reported in the ClinicalTrials.gov record. |
| Was a multiplicity strategy reported? | No multiplicity strategy is reported in the ClinicalTrials.gov record. |
| Were missing-data or imputation methods reported? | No missing-data or imputation method is reported in the ClinicalTrials.gov record. |
| Were Bayesian methods reported? | No Bayesian method is reported. The normalized statistical method is the log-rank test. |
16. Limitations
- Registry-level detail: the ClinicalTrials.gov record contains the primary and secondary endpoint definitions and posted statistical results, but several methodological details that could affect a full statistical reconstruction are not provided.
- Event-count analysis: the DFS number-of-events endpoint is listed as a primary endpoint, but the registry-reported statistical-analyses data does not provide a formal comparison for that measure.
- Secondary endpoint status: the RFS and OS analyses are explicitly described in the ClinicalTrials.gov record as descriptive analyses. Their p-values should therefore not be interpreted as though they were independently assigned confirmatory hypotheses.
- Hazard-ratio interpretation: the HR is a relative time-to-event measure and should not be translated directly into an absolute reduction in event probability.
- Proportional-hazards considerations: the ClinicalTrials.gov record identifies the hazard ratio and log-rank test but does not provide diagnostics demonstrating that proportional hazards holds throughout follow-up.
- Safety comparison: serious adverse events are reported for three listed categories, but the ClinicalTrials.gov record does not provide a formal statistical comparison or enough information to justify collapsing the two 5-FU/LV categories.
- Unreported design features: the ClinicalTrials.gov record does not report crossover, interim monitoring, multiplicity adjustment, missing-data imputation, stratification factors, or Bayesian methods. These should not be inferred from the existence of the survival analyses.
17. Why This Trial Matters Statistically
NO16968 is a useful teaching example because the registry data contains a compact but informative set of randomized-trial survival analyses. It demonstrates how a primary endpoint can be represented as a time-to-event outcome, how a log-rank test and hazard ratio work together, and why secondary endpoint p-values require context about their prespecified role.
| Concept | How it appears in NO16968 |
|---|---|
| Randomization | The trial is registered as randomized with a parallel design. |
| Intention-to-treat analysis | All three posted efficacy analyses use the ITT population. |
| Time-to-event endpoint | DFS, RFS, and OS are registered as time-to-event outcomes. |
| Log-rank test | Used for the posted DFS, RFS, and OS comparisons. |
| Hazard ratio | Reported as the effect measure for all three formal statistical analyses. |
| Confidence interval | Each posted HR has a two-sided 95% confidence interval. |
| Superiority testing | The primary DFS time-to-event analysis is identified as a superiority hypothesis. |
| Descriptive secondary analyses | RFS and OS are explicitly labeled descriptive analyses in the ClinicalTrials.gov record. |
| Censoring | Endpoint definitions use dates last known to be event-free or alive. |
| Safety reporting | Serious adverse events are reported as affected participants / participants at risk by listed treatment category. |
18. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary DFS analysis reports HR 0.80 with a two-sided 95% CI of 0.69–0.93 and P = 0.0038 using a log-rank test in the ITT population. The HR is calculated as XELOX versus 5-FU/LV.
Clinical interpretation
The registry data describes a relative time-to-event difference between the randomized groups. Clinical interpretation requires keeping the endpoint definition, absolute outcomes if available, uncertainty, treatment context, and safety information distinct rather than reducing the trial to a single p-value.
The distinction matters because statistical significance and clinical magnitude are different dimensions. The DFS HR of 0.80 is the estimated relative effect, the 0.69–0.93 interval describes uncertainty around that estimate, and the p-value addresses evidence against the tested null. None of these quantities alone describes the experience of an individual participant.
19. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
20. Related Statistical Calculators
21. Sources
- ClinicalTrials.gov: NCT00069121.
- Linked publication: PubMed record, PMID 36306483. PubMed.
- Linked publication: PubMed record, PMID 26324362. PubMed.
Continue with the underlying statistical methods
Explore the survival-analysis concepts that connect directly to the endpoints and analyses reported for NO16968.
22. Record Summary
NO16968 provides a clear example of randomized clinical-trial time-to-event analysis. The primary DFS endpoint is formally analyzed in the ITT population using a log-rank test, with a hazard ratio of 0.80, a two-sided 95% CI of 0.69–0.93, and P = 0.0038. The registry also reports descriptive analyses for RFS and OS, with HRs of 0.78 and 0.83, respectively.
The most important statistical lesson is that these numbers should be interpreted together rather than independently. The hazard ratio communicates relative event rates, the confidence interval communicates uncertainty around the estimate, the p-value addresses the tested statistical hypothesis, and the endpoint definition determines exactly what clinical event the analysis represents.