← Clinical Trials
Advanced NSCLC Phase 3 Randomized NCT04685135

KRYSTAL-12: Complete Statistical Analysis of Adagrasib in Advanced Non-Small Cell Lung Cancer

An independent statistical analysis of the randomized phase 3 KRYSTAL-12 trial comparing MRTX849 (adagrasib) with docetaxel in patients with advanced non-small cell lung cancer with a KRAS G12C mutation.

Trial start: February 23, 2021  ·  Primary completion: December 30, 2023  ·  Enrollment: 453
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.

1. Trial at a Glance

KRYSTAL-12 is a randomized, parallel-group, phase 3 oncology trial evaluating MRTX849 (adagrasib) versus docetaxel in patients with advanced non-small cell lung cancer with a KRAS G12C mutation. The registry reports a time-to-event primary endpoint and a secondary binary response endpoint, with formal statistical analyses posted for both.

453
Enrolled
Randomized trial
2
Arms
Adagrasib vs docetaxel
0.58
PFS HR
95% CI 0.45–0.76
4.68
ORR Odds Ratio
95% CI 2.56–8.56
FeatureKRYSTAL-12
PhasePhase 3
PopulationPatients with advanced non-small cell lung cancer with a KRAS G12C mutation
ConditionsMetastatic Non Small Cell Lung Cancer; Advanced Non Small Cell Lung Cancer
DesignRandomized, parallel-group, unmasked treatment trial
AllocationRandomized
Primary endpoint typeTime-to-event
Primary endpointProgression-Free Survival (PFS) as Per Blinded Independent Central Review
Secondary endpoint analyzedObjective Response Rate (ORR) as Per Blinded Independent Central Review
Enrollment453
Lead sponsorMirati Therapeutics Inc.
ClinicalTrials.govNCT04685135

2. Clinical Question

The central statistical question is whether adagrasib produces a different time-to-progression-or-death profile from docetaxel in patients with advanced non-small cell lung cancer with a KRAS G12C mutation. The registered primary endpoint is progression-free survival assessed by blinded independent central review, and the posted primary analysis uses a hazard ratio from a stratified Cox proportional-hazards model.

Population

Patients with advanced non-small cell lung cancer with a KRAS G12C mutation, as described by the registered trial population.

Intervention

MRTX849 (adagrasib).

Comparator

Docetaxel.

Primary question

Does adagrasib improve progression-free survival relative to docetaxel?

3. Trial Design

01
Randomize453 participants
02
Two armsAdagrasib vs docetaxel
03
FollowTumor progression and survival
04
AssessBlinded independent central review
05
AnalyzePFS and response
Allocation
Randomized
Design model
Parallel
Masking
None
Primary purpose
Treatment
ARM 1

Adagrasib

  • MRTX849 (adagrasib)
ARM 2

Docetaxel

  • Docetaxel

The trial began on February 23, 2021, and its primary completion date is December 30, 2023. The registry classifies the study as active, not recruiting. The enrollment field reports 453 participants.

4. Analysis Populations and Randomization

The posted primary and secondary statistical analyses both specify the Intent-To-Treat Population, defined in the registry as including all participants who are randomized into the study.

Analysis populationRegistry definition / role
Intent-To-Treat PopulationIncludes all participants who are randomized into this study; used for the posted primary PFS analysis and secondary ORR analysis.

This distinction is statistically important. An intention-to-treat analysis preserves the treatment assignment created by randomization. The treatment comparison therefore remains anchored to the original randomized groups rather than being redefined according to whether participants remained on treatment or completed a particular exposure period.

Stratified analysis: the registry analysis notes identify the PFS analysis as involving a stratified analysis and state that the hazard ratio and confidence interval come from a stratified Cox proportional-hazards model. The ClinicalTrials.gov record does not identify the individual stratification factors, so none are added here.

5. Endpoints

EndpointDefinition / registry wordingTypeAnalysis
Progression-Free Survival (PFS) as Per Blinded Independent Central Review Progression-free survival is defined as the time from randomization to the date of progression or death due to any cause, whichever occurs first. Participants who are not observed to have progressed or died are censored at the date of last evaluable tumor assessment. Disease progression was assessed as per RECISIST 1.1 in the registry definition. Time-to-event Log-rank test; stratified Cox proportional-hazards model for HR and CI
Objective Response Rate (ORR) as Per Blinded Independent Central Review From randomization till death or till disease progression or initiation of follow-up anti-cancer therapy or withdrawal, as represented in the registry time-frame field. Binary Cochran-Mantel-Haenszel chi-square test; odds ratio

6. Statistical Methodology

Log-rank test for progression-free survival

The posted primary analysis uses a log-rank test to compare the time-to-event experience between adagrasib and docetaxel. The log-rank framework uses the ordering of event times and compares the observed number of events with the number expected under the null hypothesis of no treatment difference.

For a time-to-event endpoint, this approach has an important advantage over simply comparing proportions at a single time point: participants who are event-free at their last evaluable assessment can contribute information through censoring rather than being treated as if they had experienced the event.

Stratified Cox proportional-hazards model

The registry states that the reported PFS hazard ratio and confidence interval come from a stratified Cox proportional-hazards model. The hazard ratio summarizes the estimated relative event rate between the randomized treatment groups under the fitted model.

Conceptual interpretation
HR < 1  →  lower estimated instantaneous event rate in the adagrasib group

The reported HR of 0.58 is therefore a relative time-to-event measure. It is not an absolute reduction in the percentage of participants who progress or die, and it is not the probability that an individual patient benefits.

Cochran-Mantel-Haenszel test for response

For the binary ORR endpoint, the registry reports a Cochran-Mantel-Haenszel chi-square test. This family of methods is designed for comparing categorical outcomes across treatment groups while accounting for stratification. The associated effect measure reported in the analysis is an odds ratio.

Kaplan-Meier estimation

Kaplan-Meier estimation is the standard descriptive framework for displaying a time-to-event endpoint such as PFS. It estimates the probability of remaining event-free over time while accounting for right censoring. The ClinicalTrials.gov record specifically report the log-rank test and stratified Cox model; they do not provide a Kaplan-Meier estimate in the posted statistical-analysis fields.

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

Here, di is the number of events at time ti and ni is the number at risk immediately before that event time.

7. Primary Endpoint Result: Progression-Free Survival

The posted primary statistical analysis compares adagrasib with docetaxel for progression-free survival in the Intent-To-Treat Population. The endpoint is time from randomization to progression or death, whichever occurs first, with censoring at the date of last evaluable tumor assessment for participants not observed to have progressed or died.

Hazard ratio for progression or death

0.58

95% CI: 0.45–0.76   ·   P < 0.0001

Two-sided 95% confidence interval · Superiority hypothesis

Primary PFS analysisAdagrasib vs Docetaxel
Analysis populationIntent-To-Treat Population
Statistical methodLog-rank test
Effect measureHazard ratio
Estimate0.58
95% CI0.45–0.76
P-value<0.0001
Model for HR and CIStratified Cox proportional-hazards model
HypothesisSuperiority
Clinical Biostats interpretation

The estimated hazard ratio of 0.58 means that, under the stratified Cox model, the estimated instantaneous rate of progression or death in the adagrasib group was approximately 58% of that in the docetaxel group. Expressed as a simple relative-hazard interpretation, this corresponds to an estimated 42% lower hazard because 1 − 0.58 = 0.42.

The HR does not mean that 42% fewer participants progressed or died, nor does it mean that each individual participant experienced exactly a 42% reduction in risk. It is a model-based relative measure of the event rate over the analyzed follow-up.

The two-sided 95% confidence interval of 0.45–0.76 describes uncertainty around the estimated hazard ratio. Because the entire interval is below 1, the interval is consistent with a lower estimated hazard under the fitted model for adagrasib relative to docetaxel.

The P < 0.0001 value addresses evidence against the null hypothesis under the specified testing framework. It does not measure the size of the treatment effect, does not tell us the probability that the treatment works, and should not be interpreted as a percentage measure of benefit.

Finally, the Cox interpretation depends on the model and its assumptions, including the proportional-hazards framework. The reported analysis also uses censoring, stratification, and the Intent-To-Treat population, all of which are part of the statistical context for interpreting the estimate.

8. Secondary Endpoint Result: Objective Response Rate

Objective Response Rate as Per Blinded Independent Central Review is a binary endpoint. The posted analysis uses the Intent-To-Treat Population and compares adagrasib with docetaxel using a Cochran-Mantel-Haenszel chi-square test. The registry reports an odds ratio as the effect measure.

Odds ratio for objective response

4.68

95% CI: 2.56–8.56   ·   P < 0.0001

Two-sided 95% confidence interval · Superiority hypothesis

Secondary ORR analysisAdagrasib vs Docetaxel
Analysis populationIntent-To-Treat Population
Statistical methodCochran-Mantel-Haenszel chi-square test
Effect measureOdds ratio
Estimate4.68
95% CI2.56–8.56
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

An odds ratio of 4.68 means that the estimated odds of objective response were 4.68 times as high in the adagrasib group as in the docetaxel group under the reported analysis.

The odds ratio is not a risk ratio and is not the same as saying that the response probability was 4.68 times as high. When an outcome is common, odds and probabilities can differ substantially, so the odds ratio should not be translated directly into a percentage increase in response rate without the underlying response probabilities.

The 95% confidence interval of 2.56–8.56 describes uncertainty around the odds-ratio estimate. Its lower and upper limits remain above 1, so the interval is consistent with higher response odds in the adagrasib group under the reported model and testing framework.

The P < 0.0001 result provides evidence against the null comparison specified for the categorical endpoint. It is not an effect-size measure. A very small P-value can accompany either a modest or a large effect, depending on the data and sample size.

Because this is a binary endpoint, the ORR result answers a different question from PFS. PFS describes time until progression or death, whereas ORR classifies whether a participant achieved the prespecified response outcome during the registry-defined assessment period.

9. Serious Adverse Events

The ClinicalTrials.gov record reports serious adverse events by arm as affected participants divided by the corresponding at-risk population. These figures should be read as safety counts and denominators, not as measures of efficacy.

Safety groupAffected / at risk
Adagrasib149 / 298
Docetaxel50 / 140
Crossover Adagrasib20 / 44
Safety denominator matters: the ClinicalTrials.gov record distinguishes the randomized treatment arms from a crossover-adagrasib group. The denominators therefore should not be pooled or treated as though they represented a single randomized comparison. In particular, the crossover group is a distinct category in the reported safety data.

Safety interpretation also differs conceptually from the primary efficacy analysis. The PFS result is anchored to randomized assignment in the Intent-To-Treat Population, whereas serious adverse events are naturally interpreted in relation to treatment exposure and the reported at-risk denominators.

10. Crossover and Its Statistical Importance

The trial data identify a Crossover Adagrasib safety group of 20 affected participants among 44 at risk. The ClinicalTrials.gov record therefore establish that crossover is relevant to the trial's treatment-exposure structure.

From a statistical perspective, crossover can complicate the interpretation of an intention-to-treat comparison if participants assigned to one randomized arm subsequently receive treatment associated with the other arm. The ITT analysis remains defined by the original randomization, which is important for preserving the randomized comparison, but subsequent treatment can affect observed outcomes after the point of crossover.

What randomization preserves

The ITT analysis keeps participants in the groups to which they were randomized, preserving the original treatment comparison.

What crossover changes

Subsequent exposure can make the treatment histories of the randomized groups less distinct, particularly for outcomes observed after crossover.

Important limitation: the ClinicalTrials.gov record does not provide a crossover-adjusted efficacy estimate. The reported PFS HR of 0.58 should therefore be interpreted as the posted randomized comparison rather than as a crossover-adjusted treatment effect.

11. Statistical Methods Explained

Why was a log-rank test used for PFS?

PFS is a time-to-event endpoint: participants can experience progression or death at different times, while others may remain event-free at their last assessment. A log-rank test compares the event-time distributions across randomized groups while incorporating the available follow-up and censoring information.

What does a hazard ratio of 0.58 mean?

It means the fitted stratified Cox model estimates the instantaneous progression-or-death rate in the adagrasib group at about 58% of the corresponding rate in the docetaxel group. A convenient relative-hazard interpretation is a 42% lower estimated hazard, but this should not be confused with a 42% reduction in the number of participants who eventually experience the event.

Why is the confidence interval important?

The 95% confidence interval of 0.45–0.76 shows the statistical uncertainty around the HR estimate. Reporting only 0.58 would conceal how precisely the effect was estimated. The interval also provides information about whether the estimated treatment effect is compatible with a hazard ratio of 1 under the specified inferential framework.

Why does the P-value not measure treatment effect size?

The P-value addresses compatibility with the null hypothesis under the statistical model and testing procedure. It depends on both the magnitude of the observed difference and the amount of information in the data. The HR and its confidence interval are therefore the more direct quantities for describing the estimated relative treatment effect.

What does an odds ratio of 4.68 mean?

An OR of 4.68 means that the estimated odds of objective response were 4.68 times as high in the adagrasib group as in the docetaxel group. Odds are defined as probability divided by one minus probability, so the OR should not be interpreted as a 368% increase in response probability.

Why use a Cochran-Mantel-Haenszel test?

The Cochran-Mantel-Haenszel framework is useful for comparing categorical outcomes while accounting for stratification. In KRYSTAL-12, the posted ORR analysis specifically reports a Cochran-Mantel-Haenszel chi-square test, with an odds ratio as the effect measure.

Why does intention-to-treat matter?

The ITT principle analyzes participants according to randomized assignment. This maintains the comparison generated by randomization even when participants discontinue treatment, receive subsequent therapy, or otherwise have treatment histories that differ from their initial assignment.

12. Confidence Intervals and Effect Size

The two formal analyses posted for KRYSTAL-12 provide both point estimates and two-sided 95% confidence intervals. Looking at both together gives a more complete statistical description than a P-value alone.

EndpointEffect measureEstimate95% CIP-value
Progression-Free SurvivalHazard ratio0.580.45–0.76<0.0001
Objective Response RateOdds ratio4.682.56–8.56<0.0001
Two different effect measures

The PFS analysis uses a hazard ratio because PFS is a time-to-event endpoint. The ORR analysis uses an odds ratio because ORR is binary. The numerical values 0.58 and 4.68 cannot be compared as if they were the same type of statistic.

The PFS HR below 1 indicates a lower estimated hazard in the adagrasib group, while the ORR odds ratio above 1 indicates higher estimated odds of response in the adagrasib group. The direction of the two measures is therefore different because their denominators and interpretations are different.

13. Censoring and Time-to-Event Interpretation

The registered PFS definition explicitly states that participants who are not observed to have progressed or died are censored at the date of their last evaluable tumor assessment. Censoring is a fundamental feature of survival analysis because not every participant necessarily experiences the event during observed follow-up.

Event

Progression or death, whichever occurs first, is the PFS event under the registered definition.

Censoring

A participant without observed progression or death is censored at the date of the last evaluable tumor assessment.

Censoring does not mean that the participant is treated as having had no risk after that date. Instead, the analysis uses the information available up to the censoring time. The validity of survival-analysis inference depends on assumptions about the censoring mechanism and the information available over follow-up.

14. Proportional-Hazards Considerations

The primary PFS hazard ratio comes from a stratified Cox proportional-hazards model. The proportional-hazards framework provides a compact relative comparison of event hazards, but its interpretation is strongest when the relative hazard is reasonably represented by the model over the analyzed period.

This is why the HR should be considered alongside the underlying time-to-event distribution. A single hazard ratio summarizes a potentially complex pattern of events and censoring. It does not provide a complete description of when events occurred or how treatment differences may change over time.

Interpretation caution: the ClinicalTrials.gov record does not provide a formal test or diagnostic for the proportional-hazards assumption. The HR should therefore be described as the estimate from the reported stratified Cox model rather than as a guarantee that the hazard ratio was constant at every point in follow-up.

15. Multiplicity and Hypothesis Testing

The trial data identify one registered primary endpoint and a posted secondary endpoint. The primary PFS analysis and secondary ORR analysis both report superiority hypotheses, two-sided 95% confidence intervals, and P-values below 0.0001.

EndpointRoleHypothesisInference reported
Progression-Free SurvivalPrimarySuperiorityHR 0.58; 95% CI 0.45–0.76; P < 0.0001
Objective Response RateSecondarySuperiorityOR 4.68; 95% CI 2.56–8.56; P < 0.0001

Multiplicity is important whenever several hypotheses are tested because the chance of obtaining at least one apparently positive result can increase as the number of tests increases. The ClinicalTrials.gov record does not specify an alpha-allocation or multiplicity-adjustment strategy for the relationship between the primary and secondary endpoints. Accordingly, this page does not infer one.

Statistical discipline: a very small P-value for a secondary endpoint does not by itself establish that the secondary endpoint was part of a formally multiplicity-controlled confirmatory testing hierarchy. The ClinicalTrials.gov record supports reporting the posted ORR analysis, but not reconstructing an unreported multiplicity procedure.

16. What the Primary Result Does — and Does Not — Mean

Relative treatment effect

The PFS HR of 0.58 indicates a lower estimated instantaneous progression-or-death hazard in the adagrasib group under the reported stratified Cox model. The simple arithmetic interpretation is approximately a 42% lower estimated hazard, not a 42% lower absolute probability of progression or death.

Statistical precision

The 95% CI of 0.45–0.76 communicates uncertainty around the HR. It is not a range containing the individual treatment effects experienced by patients. It describes uncertainty around the estimated population-level effect under the specified statistical framework.

P-value

The reported P < 0.0001 provides evidence against the null hypothesis used for the primary analysis. It does not quantify the magnitude of the treatment effect and should not be converted into a probability that the treatment is effective.

Endpoint-specific meaning

PFS is defined by progression or death, so the HR describes the relative timing of those events. It is not a direct measure of tumor response, overall survival, or the proportion of participants who experienced a response.

17. Trial Timeline

February 23, 2021

Trial start

The registry reports the start date of the KRYSTAL-12 phase 3 trial.

December 30, 2023

Primary completion

The registry reports December 30, 2023 as the primary completion date.

Current registry status

Active, not recruiting

the ClinicalTrials.gov record classifies KRYSTAL-12 as ACTIVE_NOT_RECRUITING.

18. Why This Trial Matters Statistically

KRYSTAL-12 is a useful teaching example because its registry record connects a randomized phase 3 design with two different classes of statistical endpoint. The primary endpoint is time-to-event and is analyzed using survival-analysis methods, while the posted secondary response endpoint is binary and is analyzed using a stratified categorical-data method.

ConceptHow it appears in KRYSTAL-12
RandomizationThe trial is randomized and uses a parallel design.
Intention-to-treat analysisThe posted PFS and ORR analyses use the Intent-To-Treat Population.
Time-to-event endpointPFS is defined from randomization to progression or death.
Log-rank testUsed for the posted primary PFS comparison.
Hazard ratioThe primary PFS effect measure is HR 0.58.
Stratified Cox modelThe reported PFS HR and CI come from a stratified Cox proportional-hazards model.
Confidence intervalThe PFS HR has a two-sided 95% CI of 0.45–0.76.
Binary endpointORR is analyzed as a binary outcome.
Cochran-Mantel-Haenszel testThe posted ORR analysis uses the Cochran-Mantel-Haenszel chi-square test.
Odds ratioThe posted ORR effect measure is OR 4.68.
CrossoverThe safety data identify a Crossover Adagrasib group of 20/44.
Safety denominatorsSerious adverse events are reported separately for adagrasib, docetaxel, and crossover adagrasib.

19. Important Limitations and Interpretation Issues

20. Statistical Interpretation vs Clinical Interpretation

Statistical interpretation

The posted primary analysis reports a PFS HR of 0.58 with a two-sided 95% CI of 0.45–0.76 and P < 0.0001. The posted secondary ORR analysis reports an OR of 4.68 with a two-sided 95% CI of 2.56–8.56 and P < 0.0001.

Clinical interpretation

The statistical results describe differences in progression-free survival and objective-response odds between the randomized groups. They do not by themselves summarize every dimension of treatment benefit, treatment exposure, safety, or individual patient experience.

The distinction matters because a statistical effect estimate is tied to a particular endpoint and analysis population. The PFS HR describes relative time-to-event experience; the ORR odds ratio describes relative odds of response; and the serious-adverse-event counts describe a separate safety dimension. None of these measures can substitute for the others.

21. Related Statistical Tutorials

Learn more about the methods used in this trial:

22. Related Statistical Calculators

23. Sources

Continue through the Clinical Biostats statistical pathway

Explore the underlying survival-analysis, categorical-data, confidence-interval, and clinical-trial methods used to interpret randomized evidence.

24. Record Summary

KRYSTAL-12 provides a clear example of how different clinical-trial endpoints require different statistical tools. The randomized phase 3 design evaluates adagrasib against docetaxel, with PFS as the registered primary time-to-event endpoint and ORR as a posted secondary binary endpoint. The primary analysis uses a log-rank comparison and a stratified Cox proportional-hazards model, producing an HR of 0.58 with a two-sided 95% CI of 0.45–0.76 and P < 0.0001. The secondary ORR analysis uses a Cochran-Mantel-Haenszel chi-square test and reports an odds ratio of 4.68 with a two-sided 95% CI of 2.56–8.56 and P < 0.0001.

The most important statistical lesson is that these estimates answer different questions. The hazard ratio describes relative progression-or-death hazard over time, while the odds ratio describes relative odds of objective response. Confidence intervals quantify uncertainty around both estimates, whereas P-values address the corresponding null hypotheses. The serious-adverse-event data add a separate safety dimension, including distinct reported groups for adagrasib, docetaxel, and crossover adagrasib.

Clinical Biostats methodology: A trial-results page should separate the reported statistical evidence from educational interpretation. The goal is to explain what each endpoint and estimate means, what it does not mean, and which assumptions or design features should be considered when interpreting the result.