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-10 is a randomized, open-label, parallel-group phase 3 treatment study in patients with advanced or metastatic colorectal cancer with KRAS G12C mutation. The study compares MRTX849 with cetuximab against chemotherapy and has two primary time-to-event endpoints: overall survival and progression-free survival.
| Feature | KRYSTAL-10 |
|---|---|
| Phase | Phase 3 |
| Conditions | Advanced Colorectal Cancer; Metastatic Colorectal Cancer |
| Population | Patients with advanced colorectal cancer with KRAS G12C mutation |
| Design | Randomized, parallel-group, open-label |
| Allocation | Randomized |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 461 |
| Number of arms | 2 |
| Study start | 2021-06-24 |
| Primary completion date | 2026-01-16 |
| Lead sponsor | Mirati Therapeutics Inc. |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT04793958 |
2. Clinical Question
The central clinical question is whether treatment with MRTX849 plus cetuximab should be compared with chemotherapy using overall survival and progression-free survival as the primary measures of treatment effect in patients with advanced or metastatic colorectal cancer with KRAS G12C mutation.
Population
Patients with advanced colorectal cancer or metastatic colorectal cancer with KRAS G12C mutation.
Intervention
MRTX849 with cetuximab.
Comparator
Chemotherapy consisting of the mFOLFOX6 regimen or FOLFIRI regimen.
Primary question
How do the randomized treatment groups compare with respect to overall survival and progression-free survival over 30 months?
3. Trial Design
MRTX849 + cetuximab
- MRTX849
- Cetuximab
Chemotherapy
- mFOLFOX6 Regimen
- FOLFIRI Regimen
For statistical interpretation, the randomized parallel-group structure is important because the treatment comparison is defined at the point of randomization. The open-label nature of the study means that treatment assignment is not masked to participants or investigators, which is particularly relevant when considering endpoints that can be affected by assessment or treatment decisions.
4. Primary Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Overall Survival (OS) | Defined as time from date of randomization to date of death due to any cause. | 30 months |
| Progression-free Survival (PFS) | Defined as time from randomization until disease progression or death from any cause, whichever occurs first. | 30 months |
Both primary endpoints are time-to-event endpoints. That distinction determines the appropriate statistical framework. The analysis is not simply a comparison of proportions at a fixed date: each participant contributes follow-up time, and participants who have not yet experienced the event can contribute censored observations.
Overall survival
OS uses death due to any cause as the event. The clock begins at randomization and ends at death or the applicable censoring point.
Progression-free survival
PFS counts the first of disease progression or death from any cause as the event. It therefore incorporates both disease progression and death.
5. Why These Endpoints Are Time-to-Event Outcomes
A participant who remains alive at the end of observed follow-up has not necessarily experienced the OS event. Likewise, a participant who has neither progressed nor died during the observed PFS period has not necessarily completed the underlying disease process. Their exact event time is unknown, but their follow-up still contains statistical information.
This is the central reason that survival analysis is used. A method such as a simple comparison of the percentage of patients who have died would discard information about when events occurred and would require a fixed observation scheme that does not naturally accommodate unequal follow-up.
The survival function gives the probability that the event time T exceeds a specified time t. For OS, the event is death from any cause. For PFS, the event is disease progression or death from any cause, whichever occurs first.
6. Statistical Methodology
The registry identifies OS and PFS as the primary endpoints but does not report posted statistical analyses for either endpoint. For endpoints defined as time from randomization to an event, the standard analytical framework is survival analysis.
Kaplan-Meier estimation
The Kaplan-Meier estimator is commonly used to estimate the event-free probability over time. It accommodates right-censoring and provides an estimated survival curve rather than reducing follow-up to a single binary outcome.
Here, di is the number of events at event time ti, and ni is the number of participants at risk immediately before that time.
Hazard ratio
For a randomized comparison of two time-to-event distributions, a Cox proportional-hazards model is commonly used to estimate a hazard ratio. The hazard ratio summarizes the relative instantaneous event rate between treatment groups under the model.
An HR below 1 indicates a lower modeled instantaneous event rate in the numerator group, while an HR above 1 indicates a higher modeled instantaneous event rate.
Log-rank comparison
A log-rank test is commonly used to compare survival distributions between randomized treatment groups. It considers the ordering of event times throughout follow-up rather than evaluating only a single time point.
Confidence intervals
A confidence interval around a hazard ratio quantifies statistical uncertainty around the estimated treatment effect. A narrow interval indicates greater statistical precision than a wide interval, although precision and clinical importance are different concepts.
Analysis population
Randomization provides the basis for the primary comparative interpretation in a randomized trial. The appropriate analysis population and censoring rules should be defined in the statistical analysis plan. The registry information available here does not report those detailed analysis specifications.
7. Statistical Methods Explained
Why use Kaplan-Meier estimation for OS?
OS is explicitly defined as the time from randomization to death from any cause. Some participants may remain alive during observed follow-up, producing right-censored observations. Kaplan-Meier estimation incorporates those participants without treating them as though they experienced death at the end of follow-up.
Why is PFS analyzed differently from a simple response rate?
PFS is defined by time until disease progression or death, whichever occurs first. Two participants can both be progression-free at a particular calendar date while having very different amounts of follow-up. A time-to-event method retains that temporal information.
What does a hazard ratio describe?
A hazard ratio compares modeled instantaneous event rates between treatment groups over time. It is a relative measure of the event process. It is not a percentage of participants who benefit, a relative risk at a specific time point, or a statement that every participant experiences the same proportional change.
Why does censoring matter?
Censoring occurs when the event has not been observed by the time a participant's usable follow-up ends. Survival methods can incorporate these observations, but their validity depends on assumptions about the censoring mechanism and the analysis rules specified in advance.
Why is randomization important for OS and PFS?
Randomization establishes the treatment groups before subsequent outcomes are observed. When the randomized comparison is analyzed appropriately, differences in outcomes can be interpreted in the context of the assigned treatment rather than simply reflecting differences between self-selected treatment groups.
What assumption accompanies a Cox hazard ratio?
The conventional Cox proportional-hazards interpretation assumes that the relative hazard is reasonably stable over time, after accounting for the model structure. If hazards change materially relative to one another, a single hazard ratio may not fully describe the treatment effect throughout follow-up.
8. Planned Analysis
The registry identifies overall survival and progression-free survival as primary endpoints, each measured over a 30-month time frame. No formal statistical analyses are posted on ClinicalTrials.gov for these endpoints.
Overall Survival
OS would normally be analyzed as a time-to-event endpoint using methods such as Kaplan-Meier estimation to describe survival over time and a between-group comparison such as a log-rank test. A Cox proportional-hazards model can provide a hazard ratio and confidence interval when its modeling assumptions are appropriate.
Progression-free Survival
PFS would normally use the same general time-to-event framework, with the event defined exactly as stated in the registry: disease progression or death from any cause, whichever occurs first. Kaplan-Meier estimation can describe the PFS distribution, while a log-rank comparison and Cox model are common tools for treatment-group comparisons.
9. Interpreting a Future OS Hazard Ratio
If a future analysis reports an OS hazard ratio below 1 for MRTX849 with cetuximab versus chemotherapy, the estimate would represent a lower estimated instantaneous hazard of death in the MRTX849-plus-cetuximab group under the fitted model.
What it means: An OS HR below 1 would indicate a relative reduction in the modeled hazard of death associated with the intervention group compared with the chemotherapy group.
What it does not mean: It would not mean that the same percentage of every participant's probability of death was reduced, nor would it directly give the absolute survival probability at 30 months.
Confidence interval: The confidence interval would describe statistical uncertainty around the estimated hazard ratio. If it were wide, the point estimate would be relatively imprecise; if narrow, the estimate would be more precise.
P-value: A p-value would address evidence against the relevant null hypothesis under the prespecified testing framework. It would not measure the magnitude or clinical importance of the treatment effect.
Caution: Interpretation of a Cox HR depends on the proportional-hazards assumption and on the prespecified handling of censoring and analysis populations.
10. Interpreting a Future PFS Hazard Ratio
If a future analysis reports a PFS hazard ratio below 1 for MRTX849 with cetuximab versus chemotherapy, it would indicate a lower estimated instantaneous rate of the composite PFS event under the fitted model.
What it means: The PFS event is the first occurrence of disease progression or death from any cause. Therefore, a PFS hazard ratio summarizes the relative rate of reaching either component of that composite endpoint.
What it does not mean: It would not establish that the treatment prevents progression specifically, because death is also a PFS event.
Confidence interval: The interval would quantify uncertainty around the estimated relative event rate. It would not describe the range of individual patient outcomes.
P-value: A p-value would address the prespecified statistical hypothesis; it would not tell the reader how large or clinically meaningful the effect is.
Caution: PFS is particularly dependent on how progression is assessed, how assessments are timed, and how censoring rules are applied. The registry information does not report those detailed statistical-analysis specifications.
11. Understanding the 30-Month Time Frame
The registry assigns a 30-month time frame to both primary endpoints. A time frame does not mean that every participant must be observed for exactly 30 months. In survival analysis, participants can contribute varying amounts of follow-up, with event times recorded when events occur and observations censored according to prespecified rules.
30 months is an analysis window
The endpoint time frame identifies the period associated with the registered primary endpoint rather than defining a fixed binary outcome for every participant.
Events occur over time
OS and PFS retain information about when death or progression occurs, which is why survival-analysis methods are appropriate.
Censoring preserves partial information
A participant without an observed event can still contribute follow-up information up to the relevant censoring time.
Time matters statistically
A treatment comparison based only on the proportion of events would discard information about the timing of those events.
12. The Composite Nature of Progression-free Survival
PFS is defined as the time from randomization until disease progression or death from any cause, whichever occurs first. This creates a composite event endpoint with two possible event mechanisms.
| Possible first event | Role in PFS |
|---|---|
| Disease progression | Counts as the PFS event |
| Death from any cause | Counts as the PFS event |
| Neither event observed during follow-up | Participant may be censored according to the prespecified analysis rules |
This definition matters when interpreting a PFS result. A PFS hazard ratio is a treatment comparison for the composite endpoint, not a pure measure of tumor progression alone. If treatment affects progression and death differently, examining the components separately can provide additional context, provided those analyses are prespecified and appropriately interpreted.
13. Open-label Design and Statistical Interpretation
The registry classifies KRYSTAL-10 as having no masking. An open-label design does not invalidate randomized comparison, but it can matter for outcomes and decisions that involve judgment.
OS is comparatively objective
Death from any cause is the registered OS event, reducing the scope for subjective determination of whether the event occurred.
PFS can be more assessment-sensitive
Disease progression involves assessment of disease status, so the absence of masking can be relevant to how progression is evaluated or managed.
Randomization remains central
The randomized allocation provides the framework for comparing outcomes between treatment strategies.
Statistical analysis cannot remove every design feature
Modeling can address specified sources of variation, but an analysis does not transform an open-label study into a masked study.
14. Randomized Parallel-Group Structure
The study uses a parallel design with two randomized arms. Each participant is assigned to one treatment strategy, and the outcomes of the groups are compared over follow-up.
| Design feature | Statistical significance |
|---|---|
| Randomized | Creates treatment groups before outcome observation and supports comparative inference when analyzed according to the randomized assignment. |
| Parallel | Each participant contributes outcomes under one randomized treatment strategy rather than serving as their own randomized crossover comparison. |
| Two arms | The principal treatment comparison is between two randomized groups. |
| Open-label | Treatment assignment is not masked, which can be relevant to assessments and post-randomization decisions. |
15. Sample Size and Information
The registry reports an enrollment of 461 participants. Enrollment is not the same quantity as the number of observed deaths or progression events used in a time-to-event analysis.
For OS and PFS, statistical information depends strongly on the number and timing of events, the amount of follow-up, censoring, treatment allocation, and the underlying event rates. A trial with the same enrollment can provide substantially different statistical precision depending on how many events occur and when they occur.
The registry reports 461 enrolled participants, but it does not report posted event counts or primary-endpoint effect estimates in the information considered here.
16. Multiplicity and Confirmatory Interpretation
KRYSTAL-10 has two registered primary endpoints: OS and PFS. When a trial has multiple primary endpoints, the statistical analysis plan ordinarily specifies how the overall type I error rate is controlled across the hypotheses and whether success on one or both endpoints is required.
The ClinicalTrials.gov record identifies the two primary endpoints but does not report an alpha-allocation or multiplicity procedure. Therefore, a future p-value for OS or PFS should be interpreted in the context of the prespecified statistical analysis plan rather than automatically treated as an independent test at an unspecified significance level.
17. Confidence Intervals, P-values, and Effect Size
Three quantities commonly appear together in clinical-trial reporting, but they answer different statistical questions.
| Quantity | Primary role | What it does not provide by itself |
|---|---|---|
| Effect estimate | Describes the magnitude and direction of the treatment comparison. | It does not quantify all uncertainty around the estimate. |
| Confidence interval | Describes statistical uncertainty around the estimated effect. | It does not describe the distribution of individual patient outcomes. |
| P-value | Assesses compatibility of the observed data with a specified null hypothesis under the testing framework. | It does not measure effect size, clinical importance, or probability that a treatment hypothesis is true. |
For a time-to-event endpoint, these quantities are most informative when read together. The hazard ratio gives a relative effect measure, the confidence interval describes its statistical precision, and the p-value addresses the corresponding hypothesis test.
18. Missing Data and Censoring
Time-to-event studies use censoring rather than treating every incomplete observation as ordinary missing data. A participant may remain free of the relevant event when observation ends, so the analysis can use the known follow-up interval without pretending that an exact event time is known.
However, censoring assumptions are important. If the mechanism causing loss of follow-up is related to the unobserved event process in ways not adequately addressed by the analysis, estimates can be affected. Detailed censoring and missing-data rules are ordinarily documented in the statistical analysis plan.
19. Proportional-Hazards Assumption
The Cox proportional-hazards model is often used to estimate treatment hazard ratios. Its standard interpretation assumes that the relative hazard between groups is reasonably stable over time.
If proportional hazards is reasonable
A single hazard ratio can provide a useful summary of the relative event rate across the analyzed follow-up.
If hazards are not proportional
A single HR may compress a changing treatment effect into one number and therefore provide an incomplete description of the time-varying difference.
This is one reason survival curves and clinically meaningful time-specific survival estimates can complement a hazard ratio when results are available.
20. ClinicalTrials.gov Reporting Status
The registry lists KRYSTAL-10 as ACTIVE_NOT_RECRUITING. The study start date is 2021-06-24, and the primary completion date is 2026-01-16. The registry reports an enrollment of 461 and identifies two primary endpoints, both with a 30-month time frame.
| Registry element | Reported value |
|---|---|
| Status | ACTIVE_NOT_RECRUITING |
| Start date | 2021-06-24 |
| Primary completion date | 2026-01-16 |
| Enrollment | 461 |
| Primary endpoints | Overall Survival (OS); Progression-free Survival (PFS) |
| Primary endpoint time frame | 30 months |
| Posted statistical analyses | None |
21. Limitations
- No posted statistical analyses: the registry does not report OS or PFS effect estimates, confidence intervals, or p-values for the registered primary endpoints.
- Limited statistical detail: the registry information does not specify the complete statistical analysis plan, including detailed censoring rules, analysis populations, multiplicity procedure, or model specifications.
- Open-label design: the absence of masking can be relevant to endpoints involving disease assessment and treatment decisions.
- Time-to-event interpretation: OS and PFS depend on event timing, censoring, and follow-up rather than enrollment alone.
- Composite PFS endpoint: PFS includes both disease progression and death from any cause, so its interpretation differs from a progression-only endpoint.
- Proportional-hazards assumption: if a Cox model is used, a single hazard ratio may be less descriptive when the relative hazards vary substantially over time.
- Multiplicity: two registered primary endpoints create a need for a prespecified framework governing confirmatory inference across endpoints.
- Generalizability: the study population is specifically defined by advanced or metastatic colorectal cancer and KRAS G12C mutation, so findings from this population should not automatically be generalized to other populations.
22. Why This Trial Matters Statistically
KRYSTAL-10 is a useful statistical teaching case because its primary outcomes are both time-to-event endpoints within a randomized phase 3 comparison. That structure brings together several fundamental concepts in clinical-trial biostatistics: randomization, censoring, Kaplan-Meier estimation, hazard ratios, confidence intervals, hypothesis testing, composite endpoints, and multiplicity.
| Concept | How it appears in KRYSTAL-10 |
|---|---|
| Randomization | The registry classifies the allocation as randomized. |
| Parallel-group design | The study has two treatment arms in a parallel design. |
| Time-to-event analysis | Both primary endpoints are defined by time from randomization to an event. |
| Kaplan-Meier estimation | A standard method for describing OS and PFS distributions with censoring. |
| Hazard ratio | A standard relative treatment-effect measure for time-to-event endpoints. |
| Confidence intervals | Provide statistical uncertainty around estimated treatment effects. |
| Log-rank testing | A standard method for comparing survival distributions between treatment groups. |
| Composite endpoint | PFS combines disease progression and death from any cause. |
| Multiplicity | Two registered primary endpoints require attention to confirmatory error control. |
| Open-label design | No masking is reported, which can affect interpretation of assessment-sensitive outcomes. |
23. A Statistical Reading Guide for KRYSTAL-10
When formal OS and PFS results are reported, the most informative approach is to read the evidence in layers rather than focusing on a single p-value.
1. Start with the endpoint
Confirm whether the result concerns OS or PFS and use the registry definition to understand exactly what counts as the event.
2. Identify the effect measure
For a survival analysis, determine whether the reported effect is a hazard ratio, an absolute survival estimate, a median event time, or another measure.
3. Read the confidence interval
Assess how precisely the treatment effect has been estimated and whether the interval is compatible with materially different effects.
4. Read the p-value in context
Interpret statistical significance according to the prespecified testing framework, especially when there are multiple primary endpoints.
5. Examine the survival curves
When available, the Kaplan-Meier curves show how the treatment groups separate and whether the relative difference appears stable over time.
6. Check the assumptions
Consider censoring, analysis population, endpoint assessment, and the proportional-hazards assumption before reducing the result to one number.
24. Statistical Concepts in This Trial
The main statistical concepts raised by KRYSTAL-10 are time-to-event analysis, randomized treatment comparison, censoring, hazard ratios, confidence intervals, composite endpoints, and multiplicity. Together they illustrate why a clinical-trial result is more than a single p-value: the endpoint definition, estimand, analysis population, follow-up, model, uncertainty, and testing framework all affect interpretation.
25. Sources
- ClinicalTrials.gov: NCT04793958 — KRYSTAL-10.
- PubMed: PubMed record 39752885.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical-trial statistical analyses and the underlying biostatistical methods used to interpret randomized evidence.
26. Record Summary
KRYSTAL-10 is a randomized phase 3, parallel-group, open-label treatment study enrolling 461 participants with advanced or metastatic colorectal cancer with KRAS G12C mutation. The two treatment strategies are MRTX849 with cetuximab and chemotherapy consisting of the mFOLFOX6 regimen or FOLFIRI regimen. The registry identifies overall survival and progression-free survival as primary endpoints, each measured over 30 months.
Statistically, the central feature of the study is the use of time-to-event endpoints. OS measures time from randomization to death due to any cause, while PFS measures time from randomization to disease progression or death from any cause, whichever occurs first. Kaplan-Meier estimation, log-rank comparison, and Cox proportional-hazards modeling are standard analytical approaches for such endpoints, although the registry information does not report the formal statistical analysis specifications for KRYSTAL-10.