← Clinical Trials
Brain & CNS Tumors Phase 3 Time-to-Event Analysis NCT00626990

CATNON: Complete Statistical Analysis of Temozolomide in Anaplastic Glioma

An independent statistical analysis of the randomized phase 3 CATNON trial evaluating concomitant and adjuvant temozolomide in anaplastic glioma without 1p/19q loss of heterozygosity, with overall survival as the registered primary endpoint.

CATNON  ·  Phase 3  ·  Randomized  ·  Enrollment 751  ·  EORTC
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results presented here are restricted to the ClinicalTrials.gov record and the linked PubMed records identified in that data. 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

CATNON is a randomized, parallel-group phase 3 trial in brain and central nervous system tumors. The registry describes a four-arm trial designed to study two separate treatment questions: the effect of concomitant temozolomide (TMZ) and the effect of adjuvant TMZ. The primary endpoint is overall survival measured from the day of randomization.

751
Enrollment
Randomized trial
4
Arms
Parallel design
0.64
Adjuvant TMZ OS HR
95% CI 0.52–0.79
0.59
Adjuvant TMZ PFS HR
95% CI 0.49–0.70
FeatureCATNON
Trial nameCATNON
Brief titlePhase III Trial of Anaplastic Glioma Without 1p/19q Loss of Heterozygosity (LOH)
PhasePhase 3
ConditionBrain and Central Nervous System Tumors
DesignRandomized, parallel-group
MaskingNone
Primary purposeTreatment
Enrollment751
Arms4
Lead sponsorEuropean Organisation for Research and Treatment of Cancer - EORTC
Sponsor typeNetwork
Trial statusActive, not recruiting
Start2007-12
Primary completion2018-09-05
ClinicalTrials.govNCT00626990

2. Clinical Question

The registry describes CATNON as addressing two treatment questions within the same randomized four-arm trial: the effect of concomitant temozolomide and the effect of adjuvant temozolomide. The primary endpoint was overall survival measured from randomization to death from any cause.

Population

Patients enrolled in the phase 3 trial of anaplastic glioma without 1p/19q loss of heterozygosity, within the registry condition of brain and central nervous system tumors.

Intervention

Temozolomide was evaluated in two treatment dimensions: concomitant TMZ and adjuvant TMZ.

Comparator

The four randomized arms were combined before the two treatment questions were analyzed, allowing absence versus presence of concomitant TMZ and absence versus presence of adjuvant TMZ to be compared.

Primary question

Does concomitant TMZ or adjuvant TMZ change overall survival measured from randomization?

3. Trial Design

01
Randomize751 enrolled
02
4 armsParallel design
03
TreatmentRT ± concomitant TMZ ± adjuvant TMZ
04
Follow-upSurvival and progression
05
AnalysisCox proportional hazards
ARM · RT Alone

Radiotherapy alone

  • Radiotherapy (RT)
  • No concomitant TMZ
  • No adjuvant TMZ
ARM · RT & Concurrent CT

RT with concomitant TMZ

  • Radiotherapy
  • Concomitant TMZ
  • No adjuvant TMZ
ARM · RT + Adjuvant CT

RT with adjuvant TMZ

  • Radiotherapy
  • No concomitant TMZ
  • Adjuvant TMZ
ARM · RT & Concurrent CT + Adjuvant CT

RT with concomitant and adjuvant TMZ

  • Radiotherapy
  • Concomitant TMZ
  • Adjuvant TMZ

The registry's statistical-analysis fields make the factorial structure particularly important. Rather than treating the four arms only as four unrelated groups, the primary analyses combine the randomized arms to estimate two treatment effects: presence versus absence of concomitant TMZ, and presence versus absence of adjuvant TMZ.

Factorial interpretation: the two primary comparisons are conceptually different from a simple four-arm pairwise comparison. The registry reports the analysis as combining the four randomized arms so that the effect of each TMZ component can be estimated separately.

4. Endpoints

EndpointRegistry definitionTime frameType
Overall Survival as Measured From the Day of Randomization The duration of survival is the time interval between randomization and the date of death due to any cause. Patients not reported dead or lost to follow up will be censored at the date of the last follow up examination. From date from enrollment till the date of death (time till death is up to 10.9 years after patient enrollment in the study) Time-to-event; primary
Progression-free Survival Time from randomization till the date of disease progression or death. From randomization till the date of disease progression or death (time till death is up to 10.9 years after patient enrollment) Time-to-event; secondary

The registry lists one primary endpoint type: time-to-event. Four statistical analyses are posted: two for overall survival and two for progression-free survival. The analysis method reported for all four comparisons is Cox regression, normalized here as the Cox proportional-hazards model, with hazard ratio as the effect measure.

5. Statistical Methodology

Primary endpoint: overall survival

The primary endpoint is a time-to-event outcome. Each patient contributes a follow-up time beginning at randomization. The event is death from any cause. A patient who is alive at the last follow-up examination is censored at that date according to the registry definition.

Registry analysis framework
Overall survival = time from randomization → death from any cause

The posted analyses use Cox regression and report hazard ratios with confidence intervals. The four randomized arms were combined to evaluate the two separate TMZ questions.

Two treatment dimensions

QuestionComparisonEffect measureHypothesis
Concomitant TMZAbsence of concomitant TMZ vs presence of concomitant TMZHazard ratioSuperiority
Adjuvant TMZAbsence of adjuvant TMZ vs presence of adjuvant TMZHazard ratioSuperiority

Cox proportional-hazards model

The registry identifies the statistical method as Regression, Cox, corresponding to a Cox proportional-hazards model. This model estimates the relative hazard associated with a treatment comparison while using the observed event and censoring times.

Conceptual Cox model
h(t|X) = h0(t) exp(βX)

For a binary treatment indicator, the exponentiated treatment coefficient, exp(β), is interpreted as the hazard ratio under the proportional-hazards model.

Why time-to-event analysis is appropriate

Overall survival and progression-free survival are not ordinary continuous outcomes. Some patients may experience the event earlier than others, while others remain event-free at the end of observed follow-up. Censoring allows those incomplete observations to contribute information without treating them as if the event occurred.

A conventional comparison of mean survival time would not naturally accommodate this censoring structure. Time-to-event methods such as Kaplan-Meier estimation and Cox regression are designed specifically for this setting.

Kaplan-Meier estimation

The registry's posted analysis fields identify Cox regression as the formal method. Kaplan-Meier estimation is the standard descriptive companion for time-to-event endpoints because it estimates the probability of remaining event-free over time while accommodating right censoring. The ClinicalTrials.gov record does not provide Kaplan-Meier estimates or survival medians, so this page does not add numerical Kaplan-Meier results.

Kaplan-Meier concept
S(t) = ∏ti ≤ t (1 − di/ni)

At each observed event time, di events are counted among ni individuals at risk immediately before that time. The product gives the estimated probability of remaining event-free through time t.

Hazard ratio

The hazard ratio is the reported effect measure for every statistical analysis posted in the registry. A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the numerator group relative to the comparison group under the fitted model.

Interpretation
HR < 1  →  lower estimated hazard in the first group relative to the reference group

The hazard ratio is a relative, model-based time-to-event measure. It is not a probability, an absolute risk difference, a median-survival difference, or the percentage of patients who benefit.

6. Primary Results: Overall Survival

The registry posts two primary analyses for overall survival. Both use Cox regression and compare the four randomized arms after combining them according to the specific TMZ question.

Effect of concomitant temozolomide

Overall survival hazard ratio

0.97

99.1% CI: 0.73–1.28   ·   P = 0.76

Comparison: absence of concomitant TMZ vs presence of concomitant TMZ

FeatureReported analysis
EndpointOverall Survival as Measured From the Day of Randomization
AnalysisCox proportional-hazards model
ComparisonAbsence of Concomitant Temozolomide vs Presence of Concomitant Temozolomide
Hazard ratio0.97
Confidence interval99.1% CI 0.73–1.28
P-value0.76
HypothesisSuperiority
Clinical Biostats interpretation

The estimated hazard ratio of 0.97 is close to 1. Under the fitted Cox model, the estimated instantaneous rate of death for the first comparison group was approximately 0.97 times that of the comparison group. In relative terms, 0.97 corresponds to an estimated hazard approximately 3% lower, but that arithmetic description should not be mistaken for a 3% reduction in individual patients' probability of death.

The 99.1% confidence interval of 0.73–1.28 is relatively wide around the point estimate and includes 1. It therefore includes values compatible with a lower hazard as well as values compatible with a higher hazard. The interval describes uncertainty in the estimated model-based treatment effect; it does not describe the range of effects experienced by individual patients.

The P-value of 0.76 measures the compatibility of the observed data with the null hypothesis under the specified testing framework. It is not a measure of the size of the treatment effect. A large P-value does not prove that the two treatment strategies are identical, just as a small P-value would not by itself establish the magnitude or clinical importance of an effect.

The analysis is framed as a superiority test. It should therefore not be reinterpreted as a formal equivalence or non-inferiority analysis. The ClinicalTrials.gov record also do not report a non-inferiority margin.

Effect of adjuvant temozolomide

Overall survival hazard ratio

0.64

95% CI: 0.52–0.79   ·   P < 0.0001

Comparison: absence of adjuvant TMZ vs presence of adjuvant TMZ

FeatureReported analysis
EndpointOverall Survival as Measured From the Day of Randomization
AnalysisCox proportional-hazards model
ComparisonAbsence of Adjuvant Temozolomide vs Presence of Adjuvant Temozolomide
Hazard ratio0.64
Confidence interval95% CI 0.52–0.79
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

The estimated hazard ratio of 0.64 indicates that, under the fitted Cox model, the estimated instantaneous rate of death in the adjuvant-TMZ comparison was approximately 64% of the corresponding rate in the comparison group. Expressed as a relative hazard difference, this corresponds to an estimated 36% lower hazard, because 1 − 0.64 = 0.36.

This does not mean that adjuvant TMZ reduced every patient's probability of death by 36%, nor does it mean that 36% of patients benefited. A hazard ratio summarizes a relative time-to-event comparison under a statistical model.

The 95% confidence interval of 0.52–0.79 quantifies uncertainty around the estimated hazard ratio. Every value in the interval is below 1, so the interval is consistent with a lower estimated hazard for the adjuvant-TMZ comparison across the interval of model-compatible values. The confidence interval still does not describe individual patient outcomes.

The P-value < 0.0001 provides evidence against the null hypothesis used for the superiority comparison. It does not measure the size of the effect; the hazard ratio and its confidence interval provide that information. Interpretation also depends on the analysis population, censoring process, model assumptions, and the way the two treatment dimensions were constructed from the four randomized arms.

7. Secondary Results: Progression-Free Survival

Progression-free survival is a secondary time-to-event endpoint defined as the time from randomization until disease progression or death. The registry posts two Cox regression analyses corresponding to the same two TMZ treatment questions used for overall survival.

Effect of concomitant temozolomide

Progression-free survival hazard ratio

0.86

95% CI: 0.72–1.03   ·   P = 0.11

Comparison: absence of concomitant TMZ vs presence of concomitant TMZ

FeatureReported analysis
EndpointProgression-free Survival
Time frameFrom randomization till the date of disease progression or death
AnalysisCox proportional-hazards model
ComparisonAbsence of Concomitant TMZ vs Presence of Concomitant TMZ
Hazard ratio0.86
Confidence interval95% CI 0.72–1.03
P-value0.11
HypothesisSuperiority
Clinical Biostats interpretation

The estimated hazard ratio of 0.86 corresponds to an estimated instantaneous rate of progression or death approximately 14% lower for the first comparison group under the Cox model, because 1 − 0.86 = 0.14.

The 95% confidence interval of 0.72–1.03 includes 1. Thus, the interval contains both values compatible with a lower hazard and values slightly above 1. The interval is a more informative description of precision than the point estimate alone.

The P-value of 0.11 does not measure the magnitude of the hazard ratio. It indicates that the observed result is not sufficiently incompatible with the null hypothesis at conventional significance thresholds under the reported superiority-testing framework. It should not be converted into a probability that the treatment does or does not work.

Because progression-free survival counts either progression or death as the event, it is not interchangeable with overall survival. The two endpoints can show different estimated effects even when they are evaluated within the same randomized trial.

Effect of adjuvant temozolomide

Progression-free survival hazard ratio

0.59

95% CI: 0.49–0.70   ·   P < 0.0001

Comparison: absence of adjuvant TMZ vs presence of adjuvant TMZ

FeatureReported analysis
EndpointProgression-free Survival
Time frameFrom randomization till the date of disease progression or death
AnalysisCox proportional-hazards model
ComparisonAbsence of Adjuvant TMZ vs Presence of Adjuvant TMZ
Hazard ratio0.59
Confidence interval95% CI 0.49–0.70
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

The estimated hazard ratio of 0.59 indicates that the estimated instantaneous rate of progression or death was approximately 59% of the corresponding rate in the comparison group under the fitted Cox model. Expressed as a relative hazard difference, this corresponds to an estimated 41% lower hazard.

The 95% confidence interval of 0.49–0.70 lies entirely below 1, providing a range of model-compatible estimates that all correspond to a lower estimated hazard for the adjuvant-TMZ comparison.

The P-value < 0.0001 indicates strong evidence against the null hypothesis within the reported superiority framework. It is not an effect-size measure and should not be interpreted as the probability that the observed hazard ratio occurred by chance.

As with the overall-survival analysis, this is a model-based result. It does not imply that every patient has the same relative reduction in risk, and it relies on the assumptions underlying the Cox analysis and the trial's censoring and analysis procedures.

8. Summary of Posted Efficacy Analyses

EndpointComparisonHRCIP-valueRole
Overall survival Absence of concomitant TMZ vs presence of concomitant TMZ 0.97 99.1% CI 0.73–1.28 0.76 Primary
Overall survival Absence of adjuvant TMZ vs presence of adjuvant TMZ 0.64 95% CI 0.52–0.79 <0.0001 Primary
Progression-free survival Absence of concomitant TMZ vs presence of concomitant TMZ 0.86 95% CI 0.72–1.03 0.11 Secondary
Progression-free survival Absence of adjuvant TMZ vs presence of adjuvant TMZ 0.59 95% CI 0.49–0.70 <0.0001 Secondary

This compact table illustrates why a four-arm trial cannot be summarized adequately by a single hazard ratio. The registry reports two distinct treatment questions, each evaluated for both overall survival and progression-free survival. The estimates differ according to both the TMZ component and the endpoint being analyzed.

9. Serious Adverse Events by Arm

The registry provides serious adverse-event counts by randomized arm as affected patients divided by patients at risk. These are presented exactly as reported in the registry rather than converted into additional percentages.

Randomized armAffected / at risk
RT Alone13 / 186
RT & Concurrent CT27 / 185
RT + Adjuvant CT32 / 183
RT & Concurrent CT + Adjuvant CT32 / 185
Serious adverse events: affected patients / patients at risk
RT Alone
13 / 186
RT & Concurrent CT
27 / 185
RT + Adjuvant CT
32 / 183
RT & Concurrent CT + Adjuvant CT
32 / 185

These serious-adverse-event figures are descriptive arm-level safety data. The ClinicalTrials.gov record does not provide a formal statistical comparison of serious adverse-event rates, confidence intervals, or a prespecified safety hypothesis test, so no inferential comparison is added here.

10. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

Both overall survival and progression-free survival are time-to-event endpoints with censoring. The Cox model is designed to compare event hazards over follow-up while using the observed timing of events and censored observations. It produces a hazard ratio that summarizes the relative event rate associated with the treatment comparison under the model.

What does a hazard ratio of 0.64 mean?

A hazard ratio of 0.64 means that the fitted model estimates the instantaneous hazard in the numerator group to be 64% of that in the reference group, subject to the model assumptions. It corresponds to a 36% lower estimated hazard, not a 36% lower probability of death for every patient.

Why are the four arms combined before analysis?

The registry states that CATNON studied two questions separately: the effect of concomitant TMZ and the effect of adjuvant TMZ. Combining the four randomized arms according to each treatment dimension permits the trial to estimate those component effects rather than restricting the analysis to one selected pair of arms.

What does the confidence interval add?

The hazard ratio is a point estimate. The confidence interval adds information about statistical precision. For example, the adjuvant-TMZ overall-survival estimate is 0.64 with a 95% CI of 0.52–0.79. The interval communicates that the point estimate is not known exactly and provides a range of values compatible with the specified statistical framework.

Why doesn't the P-value measure the size of the effect?

A P-value addresses compatibility with a null hypothesis under the statistical model and testing framework. It depends on both the magnitude of the observed result and the amount of information in the data. Effect size should therefore be read from the hazard ratio and its confidence interval, not from the P-value alone.

What does censoring mean in CATNON?

For overall survival, the registry states that patients not reported dead or lost to follow-up are censored at the date of their last follow-up examination. Censoring means that the patient's exact event time is not observed during the relevant follow-up window. Time-to-event methods use the information available up to the censoring time rather than assigning an artificial event time.

What is the proportional-hazards assumption?

The Cox model is commonly interpreted under a proportional-hazards assumption: the relative hazard between comparison groups is approximately stable over time. A single hazard ratio can be less representative if the relative hazards change substantially during follow-up. The ClinicalTrials.gov record identifies the Cox model but do not provide a diagnostic assessment of this assumption.

11. Understanding the Two Primary Comparisons

The most important statistical feature of CATNON is that the trial's primary endpoint was evaluated through two treatment questions rather than one simple intervention-versus-control comparison.

Question 1
Concomitant TMZ
Absence of concomitant TMZ vs presence of concomitant TMZ.
Question 2
Adjuvant TMZ
Absence of adjuvant TMZ vs presence of adjuvant TMZ.
Primary endpoint
Overall survival
Time from randomization to death from any cause.
Secondary endpoint
Progression-free survival
Time from randomization to disease progression or death.

This structure also explains why it would be misleading to describe the trial as though every patient receiving one specific combination were compared against every patient receiving another single combination. The registry's posted analyses explicitly define the groups according to presence or absence of each TMZ component.

Interpretive caution: the four-arm randomization is important to the causal structure of the trial. The reported hazard ratios answer the registry-defined component questions; they are not estimates of every possible pairwise comparison among the four arms.

12. Confidence Intervals and Precision

The four posted analyses illustrate several different combinations of point estimates and uncertainty intervals.

AnalysisEstimateConfidence intervalRelationship to HR = 1
OS — concomitant TMZ0.9799.1% CI 0.73–1.28Includes 1
OS — adjuvant TMZ0.6495% CI 0.52–0.79Entirely below 1
PFS — concomitant TMZ0.8695% CI 0.72–1.03Includes 1
PFS — adjuvant TMZ0.5995% CI 0.49–0.70Entirely below 1

The confidence intervals also differ in their stated coverage: the concomitant-TMZ overall-survival analysis reports a 99.1% two-sided confidence interval, whereas the other three posted analyses report 95% two-sided confidence intervals. Those levels should not be silently standardized when reproducing the registry results.

Do not equate “CI includes 1” with “no effect.” A confidence interval that crosses 1 indicates uncertainty that includes the null hazard ratio under the specified confidence level and model. It does not prove that the true treatment effect is exactly 1.

13. P-Values and Superiority Testing

All four posted analyses specify a superiority hypothesis. The registry does not report a non-inferiority margin for these analyses.

ComparisonP-valueStatistical role
OS — concomitant TMZ0.76Superiority test
OS — adjuvant TMZ<0.0001Superiority test
PFS — concomitant TMZ0.11Superiority test
PFS — adjuvant TMZ<0.0001Superiority test

The two very small P-values occur for the adjuvant-TMZ comparisons of both overall survival and progression-free survival. The concomitant-TMZ analyses have larger P-values. The P-values should be interpreted together with their corresponding hazard ratios and confidence intervals rather than used as a ranking system for the four analyses.

14. Multiplicity and the Factorial Structure

CATNON raises an important statistical issue because the trial evaluates two treatment questions across two time-to-event endpoints. The ClinicalTrials.gov record identifies four posted analyses: two primary overall-survival analyses and two secondary progression-free-survival analyses.

DimensionLevels represented in posted analyses
Treatment questionConcomitant TMZ; Adjuvant TMZ
EndpointOverall survival; Progression-free survival
Hypothesis typeSuperiority
Primary endpoint typeTime-to-event
Formal modelCox proportional-hazards model

Because multiple treatment questions and endpoints are evaluated, the interpretation of statistical significance depends on the prespecified trial-analysis framework. The ClinicalTrials.gov record do not provide an alpha-allocation scheme, multiplicity-adjustment method, interim-analysis plan, or hierarchical testing procedure. This page therefore does not invent one.

What can be concluded from the ClinicalTrials.gov record: the registry identifies which analyses are primary or secondary, their superiority hypothesis, their effect estimates, confidence intervals, and P-values. It does not provide enough information here to reconstruct the complete multiplicity-control strategy.

15. Interim Analysis, Missing Data, and Stratification

Several design topics that can materially affect survival analysis require details that are not present in the registry-reported CATNON registry data.

Interim analysis

The registry-reported statistical-analysis fields do not report an interim-analysis schedule, information fractions, alpha-spending approach, or stopping boundary.

Missing data / imputation

The ClinicalTrials.gov record defines censoring for overall survival but do not report a separate missing-data or imputation strategy for the posted Cox analyses.

Stratification

The registry analysis fields identify Cox regression but do not provide stratification factors or a stratified-Cox specification.

Bayesian methods

No Bayesian analysis is identified in the registry-reported statistical-method fields. The posted method is Cox regression.

These omissions are important because statistical analysis should distinguish what the registry explicitly documents from what would merely be conventional practice. A complete statistical plan might contain additional details, but they should not be reconstructed as if they were reported results.

16. Clinical Biostats Interpretation of the Overall-Survival Findings

Concomitant TMZ

The overall-survival estimate for absence versus presence of concomitant TMZ is HR 0.97, with a 99.1% CI of 0.73–1.28 and P = 0.76. The point estimate is close to the null value of 1, while the confidence interval spans both sides of 1. The registry result therefore does not provide the same statistical pattern as the adjuvant-TMZ comparison.

Adjuvant TMZ

The overall-survival estimate for absence versus presence of adjuvant TMZ is HR 0.64, with a 95% CI of 0.52–0.79 and P < 0.0001. Under the Cox model, the estimated hazard is 36% lower for the numerator treatment comparison. The confidence interval is entirely below 1, indicating a statistically distinguishable treatment effect under the reported superiority framework.

These two primary analyses illustrate why treatment-component trials require careful specification of the estimand. The phrase “the CATNON hazard ratio” is incomplete: the registry reports separate hazard ratios for concomitant TMZ and adjuvant TMZ.

17. Clinical Biostats Interpretation of Progression-Free Survival

Concomitant TMZ

The PFS hazard ratio for absence versus presence of concomitant TMZ is 0.86 with a 95% CI of 0.72–1.03 and P = 0.11. The point estimate is below 1, but the confidence interval crosses 1. The statistical evidence is therefore different from the adjuvant-TMZ PFS comparison.

Adjuvant TMZ

The PFS hazard ratio for absence versus presence of adjuvant TMZ is 0.59, with a 95% CI of 0.49–0.70 and P < 0.0001. The estimated hazard is approximately 41% lower under the Cox model, and the confidence interval remains below 1.

Because PFS combines progression and death into one event definition, the PFS result should not be treated as merely an earlier version of the overall-survival result. It represents a different estimand and can be influenced by disease-assessment procedures as well as mortality.

18. What the Hazard Ratio Does — and Does Not — Mean

A relative time-to-event measure

A hazard ratio compares the estimated instantaneous event rates between two groups under a fitted time-to-event model. For example, an HR of 0.59 corresponds to an estimated hazard approximately 41% lower than the reference hazard, because 1 − 0.59 = 0.41.

Not an individual risk reduction

An HR of 0.59 does not mean that each patient has a 41% lower probability of progression or death. It also does not mean that 41% of patients benefit. Individual treatment effects can vary, and the hazard ratio is a population-level model-based summary.

Not a median-survival difference

The ClinicalTrials.gov record does not report median overall survival or median progression-free survival. A hazard ratio therefore cannot be converted into a median survival difference without additional information about the underlying survival distributions.

Not a cure rate

A hazard ratio does not establish how many patients are cured or how many will remain alive at a particular time. Absolute survival probabilities require survival-curve information or other time-specific estimates.

19. Censoring and the Interpretation of Long Follow-Up

The registered overall-survival definition states that patients who are not reported dead or lost to follow-up are censored at the date of their last follow-up examination. The stated time frame extends to up to 10.9 years after patient enrollment.

Long follow-up can be statistically valuable because it allows more events to accumulate and can improve the precision of time-to-event estimates. At the same time, the interpretation of censored observations depends on the assumptions underlying the censoring mechanism and the analysis.

Why censoring is not the same as survival
Censored at t = event status unknown after the last informative follow-up time

A censored patient has not been assigned a death or progression event at the censoring time. The analysis uses the patient's observed follow-up without pretending that the patient remained event-free indefinitely.

The ClinicalTrials.gov record does not provide a detailed accounting of the number or timing of censored observations. Consequently, this page does not infer censoring patterns or make assumptions about their distribution.

20. Safety and Efficacy Answer Different Questions

The efficacy analyses estimate time-to-event treatment effects, while the ClinicalTrials.gov record summarizes serious adverse events by randomized arm. These are different statistical questions and should not be collapsed into a single benefit-risk number.

DomainCATNON data reportedStatistical interpretation
EfficacyOverall survival and progression-free survival hazard ratiosInferential time-to-event analyses
SafetySerious adverse events by armDescriptive affected / at-risk counts
Effect measureHazard ratioRelative time-to-event effect
Safety effect measureNot registry-reported as an inferential estimateNo formal safety comparison added

This distinction is important because a statistically strong efficacy result does not itself quantify the frequency or severity of adverse events. Conversely, a safety count does not establish whether one treatment changes survival or progression risk.

21. Limitations and Interpretation Issues

22. Why This Trial Matters Statistically

CATNON is a useful statistical teaching case because it combines a randomized four-arm structure with two treatment-component questions and time-to-event outcomes. It demonstrates why the choice of estimand and comparison is just as important as the choice of statistical model.

ConceptHow it appears in CATNON
RandomizationRandomized phase 3 trial with 751 enrolled participants.
Factorial treatment structureFour randomized arms are combined to study concomitant TMZ and adjuvant TMZ separately.
Time-to-event endpointOverall survival is the registered primary endpoint.
Secondary time-to-event endpointProgression-free survival is analyzed from randomization to progression or death.
Cox regressionAll four posted statistical analyses use Cox regression.
Hazard ratioThe reported effect measure for every posted statistical analysis.
Confidence intervalProvides uncertainty around each hazard-ratio estimate.
P-valueTests the specified superiority hypothesis; it is not an effect-size measure.
CensoringOverall survival censors patients at last follow-up when death is not reported or the patient is not lost to follow-up.
Safety analysisSerious adverse events are reported descriptively by randomized arm.
MultiplicityTwo treatment questions are evaluated across primary and secondary time-to-event endpoints; the ClinicalTrials.gov record does not give the complete multiplicity strategy.

23. Statistical Methods in Practice

The CATNON results illustrate several principles that recur throughout clinical-trial statistics.

Define the estimand first

The meaningful question is not simply “Does TMZ work?” but which TMZ component is being evaluated, against which comparison, and for which time-to-event endpoint.

Match the model to the endpoint

OS and PFS contain both event timing and censoring, making survival-analysis methods more appropriate than ordinary comparisons of means.

Report the full effect estimate

A hazard ratio should be presented together with its confidence interval and P-value rather than using the P-value alone.

Separate evidence from interpretation

The registry reports numerical estimates. Statistical interpretation explains what those estimates mean without turning them into individual predictions.

24. Related Tutorials

Learn more about the methods used in this trial:

25. Related Calculators

26. Sources

Continue through the Clinical Biostats statistical tutorials

Explore the survival-analysis methods and statistical concepts that underpin randomized clinical-trial interpretation.

27. Record Summary

CATNON provides a clear example of how a randomized four-arm clinical trial can address more than one prespecified treatment question. The registry identifies overall survival as the primary time-to-event endpoint and reports two Cox-model analyses: one comparing absence versus presence of concomitant temozolomide and another comparing absence versus presence of adjuvant temozolomide. The corresponding overall-survival hazard ratios are 0.97 and 0.64, respectively. Progression-free survival provides the corresponding secondary analyses, with hazard ratios of 0.86 and 0.59.

The most important statistical lesson is that the numerical result cannot be separated from the comparison being made. A hazard ratio is meaningful only in relation to its endpoint, treatment contrast, analysis population, censoring structure, model, confidence interval, and hypothesis. CATNON therefore serves as a useful example of the broader principle that careful clinical-trial interpretation requires reconstructing the statistical question before interpreting the number.

Clinical Biostats methodology: This page distinguishes registry-reported results from statistical explanation. Where the registry-reported CATNON data do not provide median survival, subgroup estimates, interim-analysis rules, multiplicity procedures, imputation methods, or other analysis details, those quantities and procedures have not been invented or inferred as reported facts.