← Clinical Trials
Gastric Cancer Phase 3 Time-to-Event NCT01170663

RAINBOW: Complete Statistical Analysis of Ramucirumab Plus Paclitaxel in Gastric Cancer

An independent statistical review of the randomized, double-blind phase 3 RAINBOW trial comparing ramucirumab plus paclitaxel with placebo plus paclitaxel in metastatic gastric adenocarcinoma.

Trial start: December 2010  ·  Primary completion: July 2013  ·  Status: Completed
Scope of this record

This page separates reported trial results from statistical interpretation. Trial-specific numerical results and methodological details are restricted to the ClinicalTrials.gov record.

Registry note: 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

RAINBOW was a randomized, double-blind, parallel phase 3 trial evaluating ramucirumab plus paclitaxel versus placebo plus paclitaxel in metastatic gastric adenocarcinoma. The registry reports one primary time-to-event endpoint, overall survival, together with secondary analyses of progression-free survival, time to progressive disease, objective response, and quality of life.

665
Enrollment
Randomized trial
2
Arms
Parallel design
0.807
OS HR
95% CI 0.678–0.962
0.0169
OS P-value
Two-sided
FeatureRAINBOW
Trial nameRAINBOW
NCT identifierNCT01170663
PhasePhase 3
StatusCompleted
ConditionGastric cancer
Brief titleA Study of Paclitaxel With or Without Ramucirumab (IMC-1211B) in Metastatic Gastric Adenocarcinoma
AllocationRandomized
Design modelParallel
MaskingDouble
Primary purposeTreatment
Enrollment665
Lead sponsorEli Lilly and Company
Sponsor typeIndustry
Results postedYes

2. Clinical Question

The central question was whether adding ramucirumab to paclitaxel improved the time from randomization to death compared with placebo plus paclitaxel in the trial population with metastatic gastric adenocarcinoma.

Population

Participants enrolled in the phase 3 RAINBOW trial for metastatic gastric adenocarcinoma.

Intervention

Ramucirumab (IMC-1211B) plus paclitaxel.

Comparator

Placebo plus paclitaxel.

Primary question

Does ramucirumab plus paclitaxel improve overall survival relative to placebo plus paclitaxel?

3. Trial Design

01
Randomize665 participants
02
Two armsRamucirumab or placebo
03
PaclitaxelBackground treatment
04
Follow-upTime-to-event outcomes
05
AnalysisSurvival and response
ARM 1

Ramucirumab and Paclitaxel

  • Ramucirumab (IMC-1211B) DP
  • Paclitaxel
  • Primary efficacy comparison against placebo plus paclitaxel
ARM 2

Placebo and Paclitaxel

  • Placebo
  • Paclitaxel
  • Comparator treatment group
Allocation
Randomized allocation to one of two parallel treatment groups.
Masking
Double-masked trial design.
Primary purpose
Treatment.
Enrollment period
Start: December 2010. Primary completion: July 2013.

4. Trial Structure and Analysis Populations

The registry's primary overall-survival analysis used all participants according to the treatment group to which they were randomized. This is an intention-to-treat-style treatment assignment principle: the comparison preserves the original randomized groups rather than moving participants between efficacy groups according to treatment actually received.

Analysis featureRegistry description
Primary OS populationAll participants according to the treatment group to which they were randomized.
OS censored participantsRamucirumab plus paclitaxel: 74; placebo plus paclitaxel: 75.
PFS censored participantsRamucirumab plus paclitaxel: 51; placebo plus paclitaxel: 39.
TTP censored participantsRamucirumab plus paclitaxel: 107; placebo plus paclitaxel: 94.
ORR populationAll participants according to the treatment group to which they were randomized.
Quality-of-life populationRandomized participants with baseline and end-of-treatment Global Health Status observations.
Important registry caveat: One participant was randomized to the placebo plus paclitaxel group but received ramucirumab in error. For the intention-to-treat population, this participant remained in the placebo plus paclitaxel group. For the safety population, the participant was included in the ramucirumab group.

5. Primary Endpoint

EndpointRegistry definitionTime framePrimary analysis
Overall Survival Time (OS) OS time was measured from date of randomization to date of death from any cause. Participants who were not known to have died on or before the date of data cut-off were censored on the last date, on or before the cut-off date, on which the participant was known to be alive. Randomization up to 27.5 months Stratified log-rank test; hazard ratio

The endpoint is therefore a classic right-censored time-to-event outcome. Participants who die contribute an observed event time. Participants who remain alive through their last known follow-up contribute survival information up to that point and are then censored.

6. Secondary Endpoints

EndpointTypeTime frameReported methodEffect measure
Progression-Free Survival (PFS)Time-to-eventRandomization up to 22.2 monthsStratified log-rank testHazard ratio
Time to Progressive Disease (TTP)Time-to-eventBaseline up to 22.2 monthsStratified log-rank testHazard ratio
Percentage of Participants With CR or PR (Objective Response Rate [ORR])BinaryRandomization up to 22.2 monthsCochran-Mantel-Haenszel testOdds ratio
Change From Baseline to End of Therapy in EORTC QLQ-C30 Global Health StatusContinuousBaseline, end of therapy (up to 103 weeks)ANCOVAP-value

The statistical methods therefore span three major data structures: time-to-event outcomes, binary response outcomes, and continuous quality-of-life change. That distinction matters because the same effect measure or hypothesis test is not appropriate for all endpoint types.

7. Statistical Methodology

Stratified log-rank test

The primary OS analysis used a stratified log-rank test. The registry states that the analysis was adjusted for three stratification factors: geographic region, time-to-progression from the start of first-line therapy, and disease measurability.

A stratified log-rank comparison evaluates the separation between treatment-group survival experience while accounting for prespecified strata. Conceptually, instead of treating all event times as if they came from one homogeneous population, the comparison is constructed within strata and then combined into an overall treatment comparison.

Conceptual survival comparison
H0: survival experience is equivalent between randomized treatment groups

The reported test is a hypothesis test for a treatment-group difference in time-to-event experience. The accompanying hazard ratio supplies an estimate of the relative event hazard, while the confidence interval describes uncertainty around that estimate.

Hazard ratio

The registry reports hazard ratios for OS, PFS, and TTP. A hazard ratio compares the estimated instantaneous event rate between treatment groups over the analyzed follow-up.

Interpretation of the hazard ratio
HR < 1  →  lower estimated event hazard in the ramucirumab-plus-paclitaxel group

A hazard ratio is a relative time-to-event measure. It is not a percentage of patients who benefit, not a probability that a particular patient will survive, and not an absolute difference in survival.

Cochran-Mantel-Haenszel test

Objective response rate was analyzed with a Cochran-Mantel-Haenszel test, with adjustment for geographic region, time-to-progression from the start of first-line therapy, and disease measurability. The corresponding effect measure was an odds ratio.

This approach is useful when a binary endpoint is compared between randomized groups while accounting for prespecified stratification variables. Rather than treating the response data as a continuous measurement, the analysis compares the odds of achieving the defined response outcome.

ANCOVA

The change from baseline to end of therapy in EORTC QLQ-C30 Global Health Status was analyzed using analysis of covariance (ANCOVA). The registry identifies this as a continuous endpoint and restricts the analysis population to randomized participants who had both baseline and end-of-treatment Global Health Status observations.

ANCOVA is a linear-model framework that can compare treatment groups on a continuous outcome while accounting for covariate information. In a change-from-baseline setting, adjustment can improve precision when relevant baseline information explains variation in the outcome. The ClinicalTrials.gov record does not specify the individual covariates included in the ANCOVA model, so none are inferred here.

Covariate adjustment and stratification

The survival and response analyses explicitly reference the same three adjustment factors: geographic region, time-to-progression from the start of first-line therapy, and disease measurability. Adjustment does not change the randomized treatment assignment. Instead, it incorporates prespecified structural information into the statistical comparison.

8. Primary Result: Overall Survival

The registry reports one formal primary-endpoint analysis for overall survival. Participants were analyzed according to randomized treatment group, and the comparison used a stratified log-rank test.

Hazard ratio for overall survival

0.807

95% CI: 0.678–0.962   ·   P = 0.0169

Two-sided confidence interval  ·  Superiority hypothesis

Primary OS featureReported result
ComparisonRamucirumab plus paclitaxel vs placebo plus paclitaxel
Analysis methodStratified log-rank test
Effect measureHazard ratio
Estimate0.807
95% CI0.678–0.962
P-value0.0169
Analysis populationAll participants according to randomized treatment group
Censored participantsRamucirumab plus paclitaxel: 74; placebo plus paclitaxel: 75
AdjustmentGeographic region; time-to-progression from start of first-line therapy; disease measurability
Clinical Biostats interpretation

The reported HR of 0.807 indicates a lower estimated instantaneous rate of death in the ramucirumab-plus-paclitaxel group relative to the placebo-plus-paclitaxel group under the reported time-to-event analysis. Expressed as a simple relative interpretation, 0.807 corresponds to an estimated hazard that is approximately 19.3% lower than the comparator hazard.

The HR does not mean that 19.3% of patients benefited, that each individual patient experienced exactly a 19.3% reduction in mortality risk, or that the absolute probability of death was reduced by 19.3 percentage points. It is a relative time-to-event measure.

The 95% CI of 0.678–0.962 describes statistical uncertainty around the estimated hazard ratio. Its width reflects the precision of the estimate; it should not be interpreted as saying that individual patients experience hazard ratios somewhere inside this interval.

The P-value of 0.0169 addresses the statistical evidence against the null hypothesis under the specified analysis. It does not measure the magnitude of the treatment effect. Effect magnitude is described by the hazard ratio, while its uncertainty is described by the confidence interval.

The analysis is also subject to the usual considerations for time-to-event methods, including censoring and the interpretation of a single hazard ratio over follow-up. The registry identifies a stratified log-rank test and does not provide enough information here to independently assess whether a proportional-hazards assumption holds.

9. Secondary Result: Progression-Free Survival

Progression-free survival was evaluated from randomization through 22.2 months using a stratified log-rank test. The analysis included all participants according to randomized treatment group.

Hazard ratio for progression-free survival

0.635

95% CI: 0.536–0.752   ·   P < 0.0001

Two-sided confidence interval  ·  Superiority hypothesis

PFS featureReported result
ComparisonRamucirumab plus paclitaxel vs placebo plus paclitaxel
Time frameRandomization up to 22.2 months
Analysis methodStratified log-rank test
Effect measureHazard ratio
Estimate0.635
95% CI0.536–0.752
P-value<0.0001
Censored participantsRamucirumab plus paclitaxel: 51; placebo plus paclitaxel: 39
AdjustmentGeographic region; time-to-progression from start of first-line therapy; disease measurability
Clinical Biostats interpretation

The PFS hazard ratio of 0.635 indicates a lower estimated instantaneous rate of the PFS event in the ramucirumab-plus-paclitaxel group relative to placebo plus paclitaxel. As a simple relative interpretation, 0.635 corresponds to an estimated hazard approximately 36.5% lower than the comparator hazard.

The estimate is not a median PFS, an absolute probability of remaining progression-free, or a guarantee that an individual patient experiences a 36.5% reduction in risk. Those quantities require different information.

The 95% CI of 0.536–0.752 quantifies uncertainty around the estimated hazard ratio. The relatively narrow interval compared with the distance from 1.0 indicates that the reported estimate is more precise than it would be with a very wide interval, but precision and clinical importance remain distinct concepts.

The P-value of <0.0001 indicates strong statistical evidence against the null hypothesis used for this comparison. It is not an effect-size measure and should not be used to compare the magnitude of the PFS and OS effects.

10. Secondary Result: Time to Progressive Disease

Time to progressive disease was evaluated from baseline through 22.2 months. The registry reports a stratified log-rank analysis adjusted for geographic region, time-to-progression from the start of first-line therapy, and disease measurability.

Hazard ratio for time to progressive disease

0.596

95% CI: 0.494–0.720   ·   P < 0.0001

Two-sided confidence interval  ·  Superiority hypothesis

TTP featureReported result
ComparisonRamucirumab plus paclitaxel vs placebo plus paclitaxel
Time frameBaseline up to 22.2 months
Analysis methodStratified log-rank test
Effect measureHazard ratio
Estimate0.596
95% CI0.494–0.720
P-value<0.0001
Censored participantsRamucirumab plus paclitaxel: 107; placebo plus paclitaxel: 94
Clinical Biostats interpretation

An HR of 0.596 corresponds to an estimated instantaneous event hazard approximately 40.4% lower in the ramucirumab-plus-paclitaxel group than in the placebo-plus-paclitaxel group under the reported analysis.

The confidence interval of 0.494–0.720 communicates uncertainty around the estimate. It does not provide a range for individual patient outcomes.

The P-value of <0.0001 describes statistical evidence under the specified hypothesis test. It does not establish how large or clinically important the effect is on its own.

11. Secondary Result: Objective Response Rate

The registry defines this endpoint as the percentage of participants with complete response or partial response. It was analyzed as a binary endpoint using the Cochran-Mantel-Haenszel test.

Odds ratio for objective response

2.14

95% CI: 1.45–3.16   ·   P = 0.0001

Two-sided confidence interval  ·  Superiority hypothesis

ORR featureReported result
EndpointPercentage of participants with CR or PR
Time frameRandomization up to 22.2 months
Analysis methodCochran-Mantel-Haenszel test
Effect measureOdds ratio
Estimate2.14
95% CI1.45–3.16
P-value0.0001
AdjustmentGeographic region; time-to-progression from start of first-line therapy; disease measurability
Clinical Biostats interpretation

An odds ratio of 2.14 means that the estimated odds of achieving a complete or partial response were 2.14 times as high in the ramucirumab-plus-paclitaxel group as in the placebo-plus-paclitaxel group under the reported analysis.

An odds ratio is not the same as a risk ratio. An OR of 2.14 does not mean that 2.14 times as many patients responded, nor does it directly state the difference in response percentages between groups.

The 95% CI of 1.45–3.16 describes uncertainty around the odds-ratio estimate. The P-value of 0.0001 addresses statistical evidence against the specified null hypothesis; it does not quantify the magnitude of response improvement.

12. Secondary Result: Quality of Life

The registry reports a continuous endpoint measuring change from baseline to end of therapy in EORTC QLQ-C30 Global Health Status. The time frame was baseline to end of therapy, up to 103 weeks. The analysis population required both baseline and end-of-treatment Global Health Status observations.

Quality-of-life featureReported result
OutcomeChange from baseline to end of therapy in EORTC QLQ-C30 Global Health Status
Time frameBaseline, end of therapy (up to 103 weeks)
Outcome unitUnits on a scale
Analysis methodANCOVA
Analysis populationRandomized participants with baseline and end-of-treatment Global Health Status observations
P-value0.3973
Hypothesis typeSuperiority
Clinical Biostats interpretation

The registry reports a P-value of 0.3973 for the ANCOVA comparison. The ClinicalTrials.gov record does not report an adjusted treatment-effect estimate or confidence interval for this endpoint, so the magnitude and precision of the between-group difference cannot be reconstructed from the ClinicalTrials.gov record.

The P-value should therefore not be translated into a percentage difference or used as a measure of similarity between treatments. A nonsignificant P-value does not prove that the treatment groups are identical; it indicates that the reported analysis did not provide strong statistical evidence against its null hypothesis at the observed result.

13. Statistical Results at a Glance

EndpointMethodEffect95% CIP-value
Overall Survival — primaryStratified log-rankHR 0.8070.678–0.9620.0169
Progression-Free SurvivalStratified log-rankHR 0.6350.536–0.752<0.0001
Time to Progressive DiseaseStratified log-rankHR 0.5960.494–0.720<0.0001
Objective Response RateCochran-Mantel-HaenszelOR 2.141.45–3.160.0001
Global Health Status changeANCOVANot reportedNot reported0.3973

This table illustrates why clinical-trial results should not be reduced to a single P-value. The endpoints have different definitions, different statistical structures, and different effect measures. The OS result is the primary endpoint; PFS, TTP, ORR, and Global Health Status are secondary outcomes.

14. How the Time-to-Event Analyses Work

Right censoring

Survival analysis must accommodate participants whose event has not been observed during the available follow-up. For OS, the registry explicitly states that participants not known to have died by the data cut-off were censored at their last known date alive on or before the cut-off.

Conceptual contribution of a censored participant
Observed information = time from randomization to last known follow-up

Censoring does not mean that an event occurred at the censoring time. It means that the analysis stops observing that participant's event status at that point under the registry's specified rule.

Why Kaplan-Meier estimation is relevant

For time-to-event endpoints such as OS and PFS, Kaplan-Meier estimation is the standard descriptive framework for representing the probability of remaining event-free over time while accounting for right censoring. The ClinicalTrials.gov record identifies the endpoint as time-to-event and provide censored counts, but they do not provide a Kaplan-Meier curve or median survival estimates.

The absence of a reported median or survival probability at a particular time point in the ClinicalTrials.gov record means this page does not add one. The hazard ratios and confidence intervals are reported directly from the statistical analyses posted on ClinicalTrials.gov.

Why the stratified log-rank test fits these endpoints

The log-rank framework is designed for comparing time-to-event distributions between randomized groups. The stratified form allows the analysis to account for prespecified stratification factors rather than assuming that all participants belong to a single homogeneous comparison stratum.

15. Statistical Methods Explained

Why was a stratified log-rank test used?

OS, PFS, and TTP are time-to-event outcomes. A participant may experience the event at a particular time or may remain event-free through the last available observation. The stratified log-rank test is designed for comparing these survival experiences while accounting for specified strata.

What does an OS hazard ratio of 0.807 mean?

It means that the estimated instantaneous rate of death was lower in the ramucirumab-plus-paclitaxel group than in the placebo-plus-paclitaxel group under the reported analysis. Numerically, 0.807 corresponds to an approximately 19.3% lower estimated hazard relative to the comparator. It does not mean a 19.3-percentage-point improvement in survival.

Why does the confidence interval matter?

The point estimate is only one estimate from the observed data. The 95% confidence interval gives a range describing statistical uncertainty around that estimate under the analysis framework. A narrower interval generally conveys greater statistical precision than a wider interval, although precision is not the same as clinical importance.

Why does the P-value not measure effect size?

A P-value addresses evidence against a null hypothesis under a specified statistical model and testing procedure. It depends on both the observed data and the amount of information available. The effect estimate and confidence interval are therefore needed to understand magnitude and precision; the P-value alone cannot provide either.

Why was the Cochran-Mantel-Haenszel test used for response?

Objective response is binary: a participant either meets the CR-or-PR definition or does not. The Cochran-Mantel-Haenszel framework permits a comparison of such categorical outcomes while accounting for specified stratification variables. RAINBOW's registry analysis reports an odds ratio of 2.14 as its effect measure.

What does an odds ratio of 2.14 mean?

An odds ratio of 2.14 means that the estimated odds of response were 2.14 times the odds in the comparator group. Odds are not probabilities, so the OR cannot be read as saying that the response probability was 2.14 times higher without additional information about the underlying response rates.

Why was ANCOVA used for Global Health Status?

The Global Health Status outcome was continuous and was analyzed as change from baseline to end of therapy. ANCOVA is a linear-model approach appropriate for comparing continuous outcomes while allowing covariate adjustment. The ClinicalTrials.gov record does not identify the specific covariates used in the ANCOVA model, so no additional model specification is assumed here.

16. Stratification and Covariate Adjustment

The registry identifies three factors used in the adjustment of the primary and several secondary analyses:

Stratification factorWhere reported
Geographic regionOS, PFS, TTP, and ORR analyses
Time-to-progression from the start of first-line therapyOS, PFS, TTP, and ORR analyses
Disease measurabilityOS, PFS, TTP, and ORR analyses

Stratification is important because randomized clinical trials may deliberately balance treatment allocation within clinically or statistically relevant strata. Incorporating the same factors into the analysis can make the treatment comparison more closely aligned with the structure used during randomization.

Interpretation caution: adjustment does not mean that the analysis has removed every source of uncertainty or imbalance. It means that specified factors were incorporated into the reported statistical comparison. The ClinicalTrials.gov record does not identify additional covariates beyond those listed above.

17. Safety Results

The registry reports serious adverse events by randomized treatment arm using affected participants divided by participants at risk. These safety counts should be kept conceptually separate from efficacy analyses because the registry caveat specifies a different treatment assignment rule for the safety population.

Safety measureReported result
Ramucirumab and paclitaxel161 affected / 327 at risk
Placebo and paclitaxel146 affected / 329 at risk

The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or P-value for serious adverse events. Accordingly, the counts should be interpreted descriptively rather than converted into an unsupported comparative inference.

Safety-population caveat: one participant randomized to placebo plus paclitaxel received ramucirumab in error. The registry states that this participant remained in the placebo plus paclitaxel group for the ITT population but was included in the ramucirumab group for the Safety population.

18. Primary Analysis vs Secondary Analyses

RAINBOW illustrates an important principle in trial interpretation: a statistically significant result on a secondary endpoint is not automatically equivalent to a statistically significant result on the primary endpoint. Endpoint role, analysis method, multiplicity, and prespecified hypothesis structure all matter.

EndpointRoleStatistical structure
Overall SurvivalPrimaryTime-to-event; stratified log-rank; HR
Progression-Free SurvivalSecondaryTime-to-event; stratified log-rank; HR
Time to Progressive DiseaseSecondaryTime-to-event; stratified log-rank; HR
Objective Response RateSecondaryBinary; Cochran-Mantel-Haenszel; OR
Global Health Status changeSecondaryContinuous; ANCOVA; P-value reported

The ClinicalTrials.gov record identifies the primary endpoint as OS and the hypothesis type as superiority. They do not provide a multiplicity-adjustment scheme or alpha-allocation procedure for the secondary endpoints. This page therefore does not infer one.

19. What the Primary Result Does — and Does Not — Establish

Effect estimate

The primary OS estimate was HR 0.807. Under the reported analysis, this represents a lower estimated instantaneous rate of death for ramucirumab plus paclitaxel compared with placebo plus paclitaxel.

Relative versus absolute effect

A hazard ratio is relative. It does not tell the reader the absolute probability of surviving to a particular time, the difference in median survival, or the number of additional survivors attributable to treatment. Those quantities require additional survival estimates that are not reported in the ClinicalTrials.gov record used here.

Confidence interval

The 95% CI of 0.678–0.962 indicates uncertainty around the reported HR. It is a statistical interval for the treatment-effect estimate, not a prediction interval for individual patients.

P-value

The P-value of 0.0169 provides evidence against the null hypothesis used for the primary comparison. It should not be treated as a measure of effect magnitude, clinical importance, or probability that the treatment hypothesis is true.

Censoring

Because OS is time-to-event data, not every participant necessarily contributes an observed death. The registry reports 74 censored participants in the ramucirumab-plus-paclitaxel group and 75 in the placebo-plus-paclitaxel group. The analysis therefore depends on the handling and assumptions associated with censored observations.

20. Comparing the Reported Effect Measures

RAINBOW uses two different relative effect measures in its reported statistical analyses: hazard ratios for time-to-event endpoints and an odds ratio for objective response.

MeasureUsed forHow to read itWhat it is not
Hazard ratioOS, PFS, TTPRelative instantaneous event hazardRisk ratio, probability, median survival, or absolute risk difference
Odds ratioORRRatio of response odds between groupsRisk ratio or direct percentage-point difference

This distinction prevents a common interpretation error: treating every number below 1 or above 1 as if it represented the same quantity. HR 0.807 and OR 2.14 are fundamentally different effect measures applied to different endpoint structures.

21. Trial Timeline

December 2010

Trial start

The registry lists December 2010 as the study start date.

July 2013

Primary completion

The registry lists July 2013 as the primary completion date.

Completed

Results posted

The registry identifies the study as completed and reports 15 outcome measures and 5 statistical analyses.

22. What the Registry Reports — and What It Does Not

Information typeStatus in the ClinicalTrials.gov record
Overall survival resultReported with HR, 95% CI, and P-value
PFS resultReported with HR, 95% CI, and P-value
TTP resultReported with HR, 95% CI, and P-value
ORR resultReported with OR, 95% CI, and P-value
Global Health Status resultReported with P-value only
Serious adverse eventsReported as affected / at risk by arm
Median OSNot reported in the ClinicalTrials.gov record
Median PFSNot reported in the ClinicalTrials.gov record
Baseline characteristicsNot reported in the ClinicalTrials.gov record
Subgroup estimatesNot reported in the ClinicalTrials.gov record
Interim analysis detailsNot reported in the ClinicalTrials.gov record
Multiplicity procedureNot reported in the ClinicalTrials.gov record
Bayesian methodsNot reported in the ClinicalTrials.gov record

Keeping these distinctions explicit is important. A trial-results page should not fill gaps in a registry record with remembered values from other sources. Where the ClinicalTrials.gov record does not report a quantity, this analysis leaves it unreported.

23. Limitations

24. Why This Trial Matters Statistically

RAINBOW is a useful teaching case because it connects several core clinical-trial concepts within one randomized phase 3 analysis. Its primary endpoint is a time-to-event outcome, while the secondary endpoints include additional survival outcomes, a binary response endpoint, and a continuous quality-of-life endpoint.

ConceptHow it appears in RAINBOW
RandomizationRandomized allocation to two parallel treatment groups
Double maskingDouble-masked phase 3 design
Time-to-event analysisOS, PFS, and TTP
Stratified log-rank testPrimary OS and secondary PFS/TTP comparisons
Hazard ratioRelative effect measure for OS, PFS, and TTP
Confidence intervalUncertainty around reported HR and OR estimates
CensoringExplicitly reported for OS, PFS, and TTP
Cochran-Mantel-Haenszel testBinary objective-response analysis
Odds ratioEffect measure for objective response
ANCOVAContinuous Global Health Status change
Covariate adjustmentSpecified adjustment for geographic region, time-to-progression, and disease measurability
Analysis populationsRandomized efficacy populations and a distinct safety-population rule

The statistical lesson is broader than the individual treatment estimate. A well-structured trial analysis requires the endpoint definition, analysis population, censoring rule, statistical test, effect measure, confidence interval, and P-value to be interpreted together.

25. Related Tutorials

Learn more about the methods used in this trial:

26. Related Calculators

27. Sources

Continue through the Clinical Biostats statistical library

Use the related tutorials and calculators to explore the survival, categorical-data, continuous-outcome, and inference methods represented in this trial.

28. Record Summary

RAINBOW provides a compact example of how several statistical methods work together in a randomized phase 3 clinical trial. The primary endpoint, overall survival, was analyzed with a stratified log-rank test and reported with a hazard ratio of 0.807, a two-sided 95% confidence interval of 0.678–0.962, and a P-value of 0.0169. Secondary analyses extended the statistical framework to progression-free survival and time to progressive disease using stratified log-rank tests, objective response using the Cochran-Mantel-Haenszel test and odds ratio, and Global Health Status change using ANCOVA.

The most important statistical interpretation is therefore not simply that individual P-values were small or large. The analysis requires attention to the endpoint definition, analysis population, censoring, stratification factors, effect measure, and confidence interval. Those elements together describe what the reported treatment comparison actually estimates.

Clinical Biostats methodology: A trial-results page should not merely repeat registry values. The goal is to reconstruct the statistical story of the trial while clearly separating reported evidence from educational interpretation and avoiding unsupported estimates where the registry does not provide them.