← Clinical Trials
Gastric Cancer Phase 3 Randomized NCT01041404

ToGA: Complete Statistical Analysis of Trastuzumab in Advanced Gastric Cancer

An independent statistical analysis of the randomized phase 3 ToGA trial comparing trastuzumab plus fluoropyrimidine/cisplatin chemotherapy with fluoropyrimidine/cisplatin alone in patients with HER2-positive advanced gastric cancer.

Trial status: Completed  ·  Enrollment: 584  ·  Primary completion: 2010-06
Scope of this record

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 ClinicalTrials.gov record.

1. Trial at a Glance

ToGA was a completed, randomized, open-label phase 3 trial evaluating trastuzumab in combination with fluoropyrimidine and cisplatin chemotherapy versus fluoropyrimidine/cisplatin alone in patients with HER2-positive advanced gastric cancer. The registry reports 584 participants, two treatment arms, two registered primary endpoints, and seven posted statistical analyses.

584
Enrollment
2 treatment arms
0.74
OS HR
95% CI 0.60–0.91
0.71
PFS HR
95% CI 0.59–0.85
0.0046
OS P-value
Two-sided log-rank
FeatureToGA
PhasePhase 3
ConditionGastric Cancer
PopulationPatients with HER2-positive advanced gastric cancer
DesignRandomized, parallel-group, open-label
AllocationRandomized
Primary purposeTreatment
Primary endpointsOverall Survival (OS) - Percentage of Participants With an Event; Overall Survival - Time to Event
Primary endpoint typeTime-to-event
Hypothesis typeSuperiority
Enrollment584.0
Trial statusCompleted
Start2005-09
Primary completion2010-06
Lead sponsorHoffmann-La Roche
Sponsor typeIndustry
ClinicalTrials.govNCT01041404

2. Clinical Question

The central statistical question was whether adding trastuzumab to fluoropyrimidine/cisplatin chemotherapy changed overall survival compared with fluoropyrimidine/cisplatin alone in patients with HER2-positive advanced gastric cancer.

Population

Patients with HER2-positive advanced gastric cancer, as identified in the trial's condition and brief title.

Intervention

Trastuzumab combined with fluoropyrimidine/cisplatin chemotherapy.

Comparator

Fluoropyrimidine/cisplatin chemotherapy alone.

Primary question

Does adding trastuzumab change overall survival relative to fluoropyrimidine/cisplatin alone?

3. Trial Design

01
Randomize584 participants
02
Arm 1FP chemotherapy
03
Arm 2Trastuzumab + FP
04
Follow-upTime-to-event outcomes
05
AnalysisFAS comparisons
Allocation
Randomized allocation to two parallel treatment groups.
Masking
None. The registry identifies the study as open-label.
Primary purpose
Treatment.
Hypothesis
Superiority. The posted analyses test whether the treatment groups differ in the specified outcomes rather than establishing non-inferiority.
COMPARATOR

Fluoropyrimidine/Cisplatin

  • Fluoropyrimidine
  • Cisplatin
  • Abbreviated in the registry as FP
EXPERIMENTAL

Trastuzumab + Fluoropyrimidine/Cisplatin

  • Trastuzumab
  • Fluoropyrimidine
  • Cisplatin
  • Abbreviated in the registry as FP + H
Design interpretation: Randomization creates the basis for comparing outcomes between the two assigned groups. Because the registry identifies masking as none, the trial was open-label. The statistical analyses therefore need to be understood in the context of a randomized but unmasked comparison.

4. Registered Primary Endpoints

The registry contains two primary endpoints, both concerning overall survival. One is expressed as the percentage of participants with an event; the other is the time to an OS event. The posted formal statistical analysis is for the time-to-event endpoint.

EndpointRegistered definitionTime frameFormal analysis posted
Overall Survival (OS) - Percentage of Participants With an Event OS was defined as the time from the date of randomization to the date of death due to any cause. Participants were censored at the last date of tumor measurement, the last date in the study drug log, or the date of last follow-up. Baseline (BL), Days 1, 8, 15, 22, 43, 64, 85, 106, 127, and every 21 days until the end of study, 1 year after the cut-off date for the 2nd interim efficacy analysis No formal statistical analysis is posted for this specific endpoint in the ClinicalTrials.gov record.
Overall Survival - Time to Event The median time, in months, from the date of randomization to the date of an OS event. Participants were censored at the last date tumor measurement, the last date in the study drug log, or the date of last follow-up. BL, Days 1, 8, 15, 22, 43, 64, 85, 106, 127, and every 21 days until the end of study, 1 year after the cut-off date for the 2nd interim efficacy analysis Yes

Why the two primary endpoint formulations matter

The registry distinguishes an event-based OS endpoint from an OS time-to-event endpoint. For a time-to-event endpoint, patients can contribute information even when they have not experienced death by the end of their observed follow-up, because their observations are censored rather than discarded.

The ClinicalTrials.gov record does not provide a separate effect estimate, confidence interval, or p-value for the primary endpoint labelled "Overall Survival (OS) - Percentage of Participants With an Event." The formal primary statistical result available in the ClinicalTrials.gov record is therefore the OS time-to-event analysis.

5. Endpoints

The registered and posted outcomes cover time-to-event, binary response, and clinical-benefit measures. The registry's statistical methods include log-rank testing for time-to-event outcomes and chi-squared testing for binary outcomes.

EndpointTypeAnalysis populationReported method
Overall Survival - Time to EventTime-to-eventFASLog-rank
Progression-Free Survival - Time to EventTime-to-eventFASLog-rank
Time to Progression - Time to EventTime-to-eventFASLog-rank
Percentage With Confirmed Complete Response or Partial ResponseBinaryFASChi-squared
Duration of ResponseTime-to-eventFAS; only participants with a CR or PR includedLog-rank
Percentage With Clinical BenefitBinaryFASChi-squared for one posted analysis
Percentage With Clinical BenefitBinaryFASOdds ratio posted; method not reported

6. Statistical Methodology

Log-rank testing for time-to-event outcomes

The registry reports the log-rank test for overall survival, progression-free survival, time to progression, and duration of response. The log-rank test compares the observed pattern of event occurrence between randomized groups over follow-up while accounting for censored observations.

Conceptual comparison
H0: the time-to-event distributions are the same between treatment groups

The log-rank test is not a measure of effect size. The hazard ratio provides a separate estimate describing the relative event hazard under the corresponding time-to-event analysis.

Hazard ratio

The primary OS analysis reports a hazard ratio of 0.74 comparing FP with trastuzumab + FP as the registry lists the groups. A hazard ratio below 1 indicates a lower estimated instantaneous event rate for the first-listed group relative to the second-listed group under the fitted time-to-event comparison.

Generic interpretation
HR = hFP(t) / hFP+H(t)

An HR of 0.74 corresponds to an estimated instantaneous hazard about 26% lower in the FP group relative to the FP + H group if interpreted directly from that group ordering. This is a relative hazard statement, not a statement that 26% of patients benefit or that survival probability is 26% higher.

Chi-squared testing for binary outcomes

The percentage of participants with confirmed complete response or partial response was analyzed using a chi-squared test. The same method was reported for one analysis of clinical benefit.

For a binary endpoint, the treatment comparison is based on the distribution of participants between outcome categories. The reported difference in response rates provides an absolute contrast, while an odds ratio provides a relative comparison on the odds scale.

Analysis population

The posted analyses identify the Full Analysis Set (FAS) as the analysis population. For duration of response, the registry adds an important qualification: only participants with a complete response or partial response were included in that analysis.

Confidence intervals

The posted effect estimates use two-sided 95% confidence intervals. For the OS hazard ratio, the interval is 0.60 to 0.91. For binary response and clinical-benefit differences, the registry also reports two-sided 95% confidence intervals.

7. Primary Result: Overall Survival

The formal primary statistical analysis posted for ToGA is the overall-survival time-to-event comparison. It used the Full Analysis Set and a two-sided log-rank test under a superiority hypothesis.

Overall Survival Hazard Ratio

0.74

95% CI: 0.60–0.91   ·   P = 0.0046

Analysis population: FAS  ·  Method: log-rank test

Primary OS analysisFPTrastuzumab + FP
ComparisonFluoropyrimidine/Cisplatin versus Trastuzumab + Fluoropyrimidine/Cisplatin
Effect measureHazard ratio
Estimate0.74
95% CI0.60–0.91
P-value0.0046
HypothesisSuperiority
Clinical Biostats interpretation

The reported hazard ratio of 0.74 is a relative time-to-event measure. Because the registry lists FP first and FP + H second, the estimate corresponds to the estimated event hazard for FP relative to trastuzumab + FP. Numerically, 0.74 is 26% below 1, so under that group ordering it represents an estimated 26% lower instantaneous event hazard for the first-listed group relative to the second-listed group.

The estimate does not mean that 26% of patients survived, that 26% of patients benefited, or that every participant experienced exactly a 26% reduction in risk. It also does not directly give an absolute survival probability or an absolute difference in survival time.

The 95% confidence interval of 0.60 to 0.91 describes the statistical uncertainty around the hazard-ratio estimate under the analysis framework. Its width provides information about precision; it is not a range describing individual patient outcomes.

The P-value of 0.0046 quantifies how incompatible the observed comparison is with the null hypothesis used for the test under the specified statistical framework. It does not measure the magnitude of the treatment effect, the probability that the treatment works, or the probability that the null hypothesis is true.

As with other hazard-ratio analyses, interpretation of a single HR requires care if the relative hazards change substantially over time. The ClinicalTrials.gov record does not report a separate assessment of the proportional-hazards assumption, so that assumption should not be treated as independently verified here.

The other registered primary endpoint

The registry also lists "Overall Survival (OS) - Percentage of Participants With an Event" as a primary endpoint. Its definition is based on death from any cause, with censoring at the last tumor measurement, last date in the study drug log, or last follow-up. The ClinicalTrials.gov record does not contain a formal statistical analysis with an estimate, confidence interval, or p-value for this endpoint, so no separate comparative result is presented here.

8. Secondary Result: Progression-Free Survival

Progression-free survival was analyzed in the Full Analysis Set using a two-sided log-rank test. The registry reports a hazard ratio as the effect measure.

Progression-Free Survival Hazard Ratio

0.71

95% CI: 0.59–0.85   ·   P = 0.0002

Analysis population: FAS  ·  Method: log-rank test

Secondary PFS analysisReported result
GroupsFluoropyrimidine/Cisplatin versus Trastuzumab + Fluoropyrimidine/Cisplatin
Effect measureHazard ratio
Estimate0.71
95% CI0.59–0.85
P-value0.0002
HypothesisSuperiority
Time frameBL, Days 43, 85, and 127, and every 21 days thereafter until disease progression or the end of study, 1 year after the cut-off date for the 2nd interim efficacy analysis
Clinical Biostats interpretation

The reported PFS hazard ratio is 0.71, with the same FP-versus-FP + H group ordering shown in the registry. On that scale, the estimated instantaneous hazard of the PFS event is 29% lower for FP than for FP + H. This is a relative hazard interpretation and should not be converted into a statement about the percentage of patients who are progression-free at a particular time.

The 95% CI of 0.59 to 0.85 indicates uncertainty around the estimated hazard ratio. The interval is narrower than the corresponding OS interval of 0.60 to 0.91, although the two endpoints concern different event definitions and should not be compared simply by interval width.

The P-value of 0.0002 is evidence against the specified null hypothesis under the reported test. It is not a measure of how large the effect is. The effect magnitude is described by the HR and its confidence interval.

The PFS time frame begins at baseline and includes scheduled assessments on Days 43, 85, and 127 and every 21 days thereafter until disease progression or the end of study, according to the registry wording reported here.

9. Secondary Result: Time to Progression

Time to progression was another time-to-event endpoint analyzed in the Full Analysis Set using a log-rank test.

Time to Progression Hazard Ratio

0.70

95% CI: 0.58–0.85   ·   P = 0.0003

Analysis population: FAS  ·  Method: log-rank test

Clinical Biostats interpretation

The HR of 0.70 represents a 30% lower estimated instantaneous hazard for the first-listed FP group relative to the second-listed FP + H group under the registry's group ordering. This is a model-based relative event measure, not an absolute reduction in the probability of progression.

The 95% CI of 0.58 to 0.85 provides the uncertainty interval reported by the registry. The P-value of 0.0003 concerns the hypothesis test; it does not quantify clinical importance or effect size.

Because time to progression and progression-free survival are distinct registered endpoints, their numerical estimates should not be treated as interchangeable even though both are time-to-event outcomes.

10. Secondary Result: Confirmed Complete or Partial Response

The registry reports a binary endpoint measuring the percentage of participants with confirmed complete response or partial response determined by RECIST. The comparison used a chi-squared test in the Full Analysis Set.

Difference in Response Rates

12.80

95% CI: 4.7–20.9   ·   P = 0.0017

Effect measure: Difference in Response Rates  ·  Method: chi-squared test

Clinical Biostats interpretation

The reported effect measure is an absolute difference in response rates, not a hazard ratio. The registry reports an estimate of 12.80, with a two-sided 95% CI of 4.7 to 20.9.

This means the response-rate comparison is expressed on an absolute percentage-point scale. It should not be interpreted as a 12.80-fold increase or as a 12.80% relative increase unless the underlying rates are explicitly used to define that different measure.

The confidence interval quantifies uncertainty around the reported difference. The P-value of 0.0017 addresses the statistical test for the categorical comparison; it does not describe the magnitude of the response-rate difference.

The registry identifies the endpoint as confirmed complete response or partial response determined by RECIST. It does not provide the two arm-specific response percentages in the ClinicalTrials.gov record, so they are not reconstructed here.

11. Secondary Result: Duration of Response

Duration of response was analyzed as a time-to-event endpoint using a log-rank test. Importantly, the registry specifies that only participants with a complete response or partial response were included in this analysis.

Duration of Response Hazard Ratio

0.54

95% CI: 0.40–0.73   ·   P < 0.0001

Analysis population: FAS; only participants with a CR or PR included

Clinical Biostats interpretation

The duration-of-response HR of 0.54 corresponds to a 46% lower estimated instantaneous event hazard for the first-listed FP group relative to the FP + H group under the registry's stated group ordering.

The important analytical qualification is the population: this analysis includes only participants who achieved a complete or partial response. It therefore answers a different question from the primary OS analysis, which uses the Full Analysis Set.

The 95% CI of 0.40 to 0.73 describes uncertainty around the response-duration hazard ratio. The P-value of < 0.0001 concerns the statistical evidence under the reported log-rank test and does not itself measure the duration difference or its clinical magnitude.

12. Secondary Results: Clinical Benefit

The registry contains two posted analyses for the percentage of participants with clinical benefit. Both use the Full Analysis Set and compare fluoropyrimidine/cisplatin with trastuzumab plus fluoropyrimidine/cisplatin.

Clinical-benefit analysisEstimate95% CIP-valueMethod
Difference in Clinical Benefit Rate9.602.4–16.90.0081Chi-squared test
Odds Ratio1.661.14–2.41Not reportedNot reported
Clinical Biostats interpretation

The first analysis expresses the treatment contrast as a difference in clinical benefit rate of 9.60, with a 95% CI of 2.4 to 16.9 and P = 0.0081 from a chi-squared test. This is an absolute difference measure.

The second analysis reports an odds ratio of 1.66, with a 95% CI of 1.14 to 2.41. The registry does not report the formal method for this particular analysis in the ClinicalTrials.gov record, and it does not provide a p-value for it. The odds ratio should therefore be kept conceptually separate from the difference in rates.

An odds ratio of 1.66 does not mean that the probability of clinical benefit was 66% higher. Odds and probabilities are different quantities. Converting an odds ratio to a risk ratio requires the underlying event rates.

13. Statistical Results at a Glance

EndpointRoleEffect95% CIP-valueMethod
Overall Survival - Time to EventPrimaryHR 0.740.60–0.910.0046Log-rank
Progression-Free Survival - Time to EventSecondaryHR 0.710.59–0.850.0002Log-rank
Time to Progression - Time to EventSecondaryHR 0.700.58–0.850.0003Log-rank
Confirmed CR or PRSecondaryDifference 12.804.7–20.90.0017Chi-squared
Duration of ResponseSecondaryHR 0.540.40–0.73<0.0001Log-rank
Clinical BenefitSecondaryDifference 9.602.4–16.90.0081Chi-squared
Clinical BenefitSecondaryOR 1.661.14–2.41Not reportedNot reported
Educational note: the ClinicalTrials.gov record provides summary estimates and confidence intervals but do not provide patient-level event and censoring data. A valid Kaplan-Meier reconstruction cannot be generated from these summary statistics alone.

14. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm. The figures are presented as affected participants divided by participants at risk.

Serious adverse eventsAffected / at risk
Fluoropyrimidine/Cisplatin (FP)81 / 290
Trastuzumab + Fluoropyrimidine/Cisplatin106 / 294
Serious adverse events by arm
FP
81 / 290
Trastuzumab + FP
106 / 294
Safety denominator matters: the registry supplies serious-adverse-event counts as affected participants divided by participants at risk. These are not the same statistical quantity as the efficacy analysis population unless the denominators are explicitly defined that way. The ClinicalTrials.gov record does not provide a formal hypothesis test or confidence interval for this safety comparison.

15. Statistical Methods Explained

Why was a log-rank test used for overall survival?

Overall survival is a time-to-event endpoint: the outcome is not simply whether death occurred, but the time from randomization to death, with censoring for participants who remain under observation without a recorded event. The log-rank test is designed to compare the survival experience of two groups across follow-up while incorporating censored observations.

What does an OS hazard ratio of 0.74 mean?

With FP listed first and FP + H second, the reported HR of 0.74 represents the estimated instantaneous event hazard for FP relative to FP + H. Numerically, it is 26% below 1. It does not mean that 26% more people survived, that survival time increased by 26%, or that each individual had exactly a 26% reduction in risk.

Why is the confidence interval important?

The point estimate is only one summary of the treatment comparison. The 95% CI of 0.60 to 0.91 shows the statistical uncertainty surrounding the OS HR under the reported analysis. A confidence interval also helps distinguish a precise estimate from a highly uncertain one; it should be interpreted alongside the effect estimate rather than replaced by the P-value.

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

The OS P-value of 0.0046 addresses the compatibility of the observed data with the null hypothesis under the specified test. It is affected by both the size of the observed difference and the amount of information available. The HR and confidence interval are the quantities that describe the estimated relative effect and its uncertainty.

Why was a chi-squared test used for response?

Confirmed complete response or partial response is a binary outcome. Each participant falls into an outcome category, and the chi-squared test provides a way to compare the distribution of those categories between treatment groups. The corresponding effect measure in the registry is a difference in response rates.

Why is duration of response analyzed in responders?

The registry explicitly defines the duration-of-response analysis population as the FAS with only participants who achieved a complete response or partial response included. Conceptually, duration of response begins with achievement of a qualifying response, so it addresses how long those responses persist rather than whether patients respond in the first place.

How should the clinical-benefit odds ratio be interpreted?

An odds ratio of 1.66 compares the odds of clinical benefit between the groups in the registry's reported analysis. Odds are not probabilities. An OR of 1.66 therefore cannot be read as a 66% increase in the probability of benefit without knowing the underlying event rates.

16. Understanding the Time-to-Event Framework

Several of the ToGA outcomes are time-to-event endpoints. This is important because conventional methods for comparing simple proportions do not fully use the information available when patients are followed for different lengths of time.

Event

For overall survival, the registered event is death due to any cause.

Censoring

For OS, participants were censored at the last tumor measurement, last date in the study drug log, or last follow-up, according to the registry definition.

Analysis

The registry reports log-rank testing for the time-to-event comparisons.

Effect measure

The main time-to-event effect measure posted for these analyses is the hazard ratio.

The distinction between an event and censoring is central to survival analysis. A participant who has not experienced the event by the end of observed follow-up is not treated as though the event could never occur. Instead, the participant contributes information up to the censoring time.

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

Here, di represents events at a particular event time and ni represents participants at risk immediately before that time. The registry-reported ToGA registry data do not contain the event-level information required to calculate a trial-specific Kaplan-Meier curve.

17. Interpreting the Direction of the Hazard Ratios

A particularly important statistical detail on this page is the ordering of the groups. The registry lists the comparisons as "Fluoropyrimidine/Cisplatin (FP) vs Trastuzumab + Fluoropyrimidine/Cisplatin (FP + H)" for the primary OS analysis, and uses similar ordering for several secondary endpoints.

EndpointGroup ordering in the ClinicalTrials.gov recordHRDirect numerical interpretation
Overall SurvivalFP vs FP + H0.74FP has an estimated instantaneous event hazard 26% lower than FP + H under this ordering.
Progression-Free SurvivalFP vs FP + H0.71FP has an estimated instantaneous event hazard 29% lower than FP + H under this ordering.
Time to ProgressionFP vs FP + H0.70FP has an estimated instantaneous event hazard 30% lower than FP + H under this ordering.
Duration of ResponseFP vs FP + H0.54FP has an estimated instantaneous event hazard 46% lower than FP + H among the response-analysis population under this ordering.
Reading treatment effects carefully: A hazard ratio is directional. If the treatment and comparator labels are reversed, the reciprocal would describe the same relative comparison from the opposite direction. For this reason, the treatment-group ordering should always be checked before translating an HR into words.

18. P-values, Confidence Intervals, and Effect Measures

ToGA provides a useful illustration of why statistical interpretation should use several quantities together rather than relying on a single number.

Effect estimate

The HR, odds ratio, or rate difference describes the estimated magnitude of the comparison on its particular scale.

Confidence interval

The 95% CI describes statistical uncertainty around the corresponding estimate under the reported framework.

P-value

The P-value addresses evidence against a specified null hypothesis. It is not an effect-size measure.

Endpoint definition

The meaning of an estimate depends on the event definition, time frame, and analysis population.

For example, the OS analysis has HR 0.74, 95% CI 0.60–0.91, and P = 0.0046. The PFS analysis has HR 0.71, 95% CI 0.59–0.85, and P = 0.0002. These values should not be interpreted as three interchangeable measures of the same quantity: OS and PFS have different event definitions.

Likewise, the response-rate difference of 12.80 and the clinical-benefit odds ratio of 1.66 are on different scales. An absolute difference and an odds ratio cannot be compared numerically as though they were the same type of effect measure.

19. Trial Timeline

2005-09

Trial start

The registry identifies September 2005 as the study start.

Phase 3

Randomized treatment comparison

The completed phase 3 trial used randomized, parallel allocation to two treatment arms without masking.

2010-06

Primary completion

The registry identifies June 2010 as the primary completion date.

Completed

Results posted

The registry reports results posted, with 19 outcome measures and 7 statistical analyses in the ClinicalTrials.gov record.

20. Limitations

21. Why This Trial Matters Statistically

ToGA is a useful teaching case because it combines randomized treatment assignment with multiple types of statistical endpoints. The primary outcome is a time-to-event endpoint analyzed with a log-rank test and summarized using a hazard ratio, while several secondary outcomes demonstrate categorical-data analysis and response-duration analysis.

ConceptHow it appears in ToGA
RandomizationThe registry identifies a randomized, parallel-group phase 3 design.
Open-label designMasking is listed as none.
Time-to-event analysisOS, PFS, time to progression, and duration of response are time-to-event endpoints.
Log-rank testUsed for the posted OS, PFS, time-to-progression, and duration-of-response analyses.
Hazard ratioUsed for the primary OS analysis and several secondary time-to-event analyses.
Confidence intervalsTwo-sided 95% intervals accompany the reported effect estimates.
Chi-squared testUsed for the confirmed CR/PR endpoint and one clinical-benefit analysis.
Odds ratioPosted as an effect measure for a clinical-benefit analysis.
Analysis populationsMost posted analyses use FAS; duration of response restricts the population to CR/PR responders.
Superiority hypothesisThe posted analyses identify superiority as the hypothesis type.

22. Overall Statistical Interpretation

What the reported evidence shows

The ClinicalTrials.gov record reports a primary OS hazard ratio of 0.74 with a two-sided 95% CI of 0.60 to 0.91 and P = 0.0046, based on a log-rank analysis in the FAS. The registry also reports secondary hazard-ratio analyses for PFS, time to progression, and duration of response, together with categorical analyses for confirmed response and clinical benefit.

What the statistics do not establish by themselves

These summary statistics do not establish individual patient benefit, do not provide the complete distribution of survival times, and do not permit reconstruction of a Kaplan-Meier curve. They also do not provide the underlying arm-specific response rates for the reported response-rate difference or a formal comparative safety test.

Why several measures are needed

The OS hazard ratio describes relative time-to-event experience; the response-rate difference describes an absolute categorical contrast; the clinical-benefit odds ratio describes an odds-scale comparison. Each answers a different statistical question. A careful interpretation therefore preserves the endpoint definition, effect measure, confidence interval, P-value, and analysis population rather than reducing the trial to a single number.

23. Related Tutorials

Learn more about the methods used in this trial:

24. Related Calculators

25. Sources

Continue through the Clinical Biostats statistical pathway

Explore the statistical concepts behind randomized clinical trials, survival analysis, categorical endpoints, effect measures, and confidence intervals.

26. Record Summary

The ToGA registry record provides a compact example of how a randomized phase 3 oncology trial can be represented statistically across several endpoint types. The principal formal analysis is an OS time-to-event comparison using a log-rank test, with a reported hazard ratio of 0.74, 95% CI 0.60–0.91, and P = 0.0046. Secondary analyses extend the statistical framework to PFS, time to progression, duration of response, confirmed response, and clinical benefit.

The most important interpretive discipline is to keep each estimate attached to its endpoint definition and analysis population. Hazard ratios describe relative time-to-event hazards; rate differences describe absolute categorical contrasts; odds ratios describe relative odds. Confidence intervals quantify uncertainty around those estimates, while P-values address the corresponding hypothesis tests rather than effect magnitude.

Clinical Biostats methodology: A trial-results page should not merely repeat a registry summary. The goal is to reconstruct the statistical structure of the trial, distinguish primary from secondary analyses, explain what each effect measure means, and identify the limitations that matter when interpreting the reported evidence.