This page separates information posted in the ClinicalTrials.gov record from statistical interpretation. The ClinicalTrials.gov record contains posted outcome measures and safety results but no formal statistical analyses. Accordingly, no hazard ratio, confidence interval, p-value, median PFS, or other comparative efficacy estimate is presented here.
1. Trial at a Glance
COLUMBUS was a randomized, parallel-group, unmasked phase 3 oncology trial with 921 enrolled participants. The trial compared LGX818 plus MEK162 with vemurafenib and LGX818 monotherapy in BRAF mutant melanoma and registered two primary time-to-event endpoints based on progression-free survival assessed by blinded independent review.
| Feature | COLUMBUS |
|---|---|
| Trial name | COLUMBUS |
| NCT identifier | NCT01909453 |
| Phase | Phase 3 |
| Condition | Melanoma |
| Trial population described in the title | BRAF mutant melanoma |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 921 |
| Arms | 4 |
| Interventions | LGX818; MEK162; vemurafenib |
| Lead sponsor | Pfizer |
| Sponsor type | Industry |
| Status | Completed |
2. Clinical Question
The registry describes a study comparing combination LGX818 plus MEK162 with vemurafenib and LGX818 monotherapy in BRAF mutant melanoma. The two registered primary endpoints focus on progression-free survival for the Part 1 Combo 450 Group compared separately with the vemurafenib group and the LGX818 group.
Population
the ClinicalTrials.gov record identifies the condition as melanoma and the study title specifies BRAF mutant melanoma.
Combination intervention
LGX818 plus MEK162. The registry safety data identify Part 1 LGX818 450 mg QD + MEK162 45 mg BI and Part 2 LGX818 300 mg QD + MEK162 45 mg B as safety cohorts.
Comparator groups
The registered primary endpoints compare the Combo 450 Group with the vemurafenib group and with the LGX818 group.
Primary question
How does progression-free survival by blinded independent review compare for the Combo 450 Group versus the vemurafenib group and versus the LGX818 group?
3. Trial Design
Participants were allocated using a randomized design.
The registry identifies the design model as parallel rather than crossover or factorial.
The registry identifies the trial as having no masking.
The primary purpose is listed as treatment.
The combination of randomization and a parallel-group structure is central to the causal interpretation of the trial. Randomization creates the framework for comparing outcomes across treatment assignments, while the parallel design means participants remain in their assigned treatment pathway rather than switching according to a crossover design.
Study treatment cohorts represented in the safety data
LGX818 450 mg QD + MEK162 45 mg BI
- Serious adverse events: 84/192 affected/at risk.
- This cohort is the registry's "Combo 450 Group" referenced in both primary PFS comparisons.
LGX818 300 mg
- Serious adverse events: 71/192 affected/at risk.
- This is the LGX818 group named in the second primary PFS comparison.
Vemurafenib 960 mg BID
- Serious adverse events: 78/186 affected/at risk.
- This is the vemurafenib group named in the first primary PFS comparison.
LGX818 300 mg QD + MEK162 45 mg B
- Serious adverse events: 98/257 affected/at risk.
- The ClinicalTrials.gov record identifies this as a separate Part 2 combination cohort.
4. Enrollment, Timing, and Trial Status
| Trial milestone | Registry value |
|---|---|
| Start | 2013-09-12 |
| Primary completion | 2016-11-09 |
| Status | COMPLETED |
| Enrollment | 921 |
| Results posted | Yes |
| Outcome measures posted | 37 |
The distinction between results posted and statistical analyses posted is important for interpreting this record. The ClinicalTrials.gov record indicates that results and 37 outcome measures have been posted, but no formal statistical analyses were posted. Consequently, the ClinicalTrials.gov record does not provide the formal effect estimates, confidence intervals, or p-values needed to describe a comparative efficacy result.
5. Primary Endpoints
| Registered primary endpoint | Time frame | Endpoint type | Results posted |
|---|---|---|---|
| Part 1: Progression Free Survival (PFS) by BIRC in Combo 450 Group as Compared to Vemurafenib Group | From randomization until documented disease progression (PD), initiation of new anti-cancer therapy, censoring date or death, whichever occurred first (up to 29 months) | Time-to-event | Yes |
| Part 1: PFS by BIRC in Combo 450 Group as Compared to LGX818 Group | From randomization until documented PD, initiation of new anti-cancer therapy, censoring date or death, whichever occurred first (up to 29 months), excluding Part 1: LGX818 300 mg group; up to 35 months for Part 1: LGX 300 mg group | Time-to-event | Yes |
Primary endpoint definition
The registry defines PFS as the time from the date of randomization to the date of the first documented disease progression or death due to any cause, whichever occurs first. PFS was determined from tumor assessment using RECIST version 1.1 criteria together with blinded independent review and survival information.
The registry definition further states that when a participant did not have an event at the analysis cut-off or at the start of new anti-cancer therapy, data were censored. This makes PFS a classic time-to-event endpoint in which both observed events and appropriately handled censored observations contribute to the analysis.
6. Planned Analysis: Primary PFS Endpoints
The registry indicates that results were posted for both primary PFS endpoints, but no formal statistical analyses were posted. Therefore, this page does not report a hazard ratio, confidence interval, p-value, median PFS, or other comparative efficacy estimate.
What the registry establishes
Combo 450 Group versus Vemurafenib Group
Combo 450 Group versus LGX818 Group
How a PFS endpoint of this type is normally analyzed
For a randomized time-to-event endpoint such as PFS, a typical analysis begins by estimating the distribution of time to progression or death with the Kaplan-Meier method. A treatment-group comparison can then be made using a log-rank test, while a Cox proportional-hazards model can estimate a hazard ratio and its confidence interval.
The appropriate analysis also requires a prespecified censoring rule. That rule is particularly important for PFS because participants may reach the end of observation without progression or death, may start another anti-cancer therapy, or may otherwise leave the observation process before an event is recorded.
The registry definition makes progression and death competing event types for the composite PFS endpoint: whichever occurs first determines the observed event time.
What is not available in the registry-reported statistical record
- No hazard ratio is provided in the posted results.
- No confidence interval is provided in the posted results.
- No p-value is provided in the posted results.
- No median PFS is provided in the posted results.
- No formal comparative statistical result is provided for either primary endpoint in the ClinicalTrials.gov record.
7. Clinical Biostats Interpretation of the Primary Endpoint Structure
The trial registered two primary PFS comparisons involving the same Combo 450 Group: one against vemurafenib and one against LGX818. These are distinct randomized comparisons, but they belong to the same overall confirmatory trial framework. The ClinicalTrials.gov record does not specify an alpha-allocation or multiplicity procedure for these comparisons, so no claim about formal type I error control should be added.
If a Cox model were used, its hazard ratio would summarize the estimated relative instantaneous rate of progression or death between the two randomized groups. A hazard ratio below 1 would indicate a lower estimated event rate for the numerator group relative to the comparator under the model. It would not mean that the same percentage of patients avoided progression, nor would it directly provide a difference in median PFS.
A 95% confidence interval around a hazard ratio would quantify statistical uncertainty around the estimated treatment effect under the specified model and sampling framework. A narrow interval would indicate greater precision than a wide interval. The interval would not describe the range of treatment effects experienced by individual participants.
A p-value would address compatibility of the observed data with a specified null hypothesis under the chosen statistical model. It would not tell us how large or clinically important the treatment effect is. Effect estimates and confidence intervals are therefore necessary to understand magnitude and precision.
PFS analysis depends on rules for determining when participants are censored. Different censoring rules can change which observations are treated as events and which contribute follow-up without an observed event. For that reason, the registry's explicit PFS definition is a core part of the statistical specification rather than administrative detail.
8. Statistical Methodology
Time-to-event analysis
Both registered primary endpoints are classified as time-to-event outcomes. Unlike a simple binary endpoint, a time-to-event analysis uses both whether an event occurred and the amount of observed follow-up before the event or censoring.
Here, the survival function represents the probability of remaining event-free beyond time t. For PFS, the event is the first documented progression or death according to the registered definition.
Kaplan-Meier estimation
The Kaplan-Meier estimator is commonly used to estimate PFS over time while accommodating right-censored observations. Each participant contributes information until the first event or censoring time. The estimator then combines the observed event information across successive event times.
Here, di represents the number of events at an event time and ni the number at risk immediately beforehand.
Cox proportional-hazards modeling
A Cox proportional-hazards model is a standard approach for estimating a relative hazard between treatment groups. The model is attractive for clinical trials because it does not require specifying the baseline hazard function parametrically.
The hazard ratio is a relative time-to-event measure. It is not a relative risk, an absolute risk difference, or the percentage of participants who benefit.
Log-rank comparison
A log-rank test is commonly used to compare time-to-event distributions between randomized groups. It takes account of the ordering of event times and censored observations rather than reducing each participant to a simple event/no-event indicator.
Blinded independent review
The registered endpoints specify PFS by BIRC. The registry-reported definition states that tumor assessments were based on RECIST version 1.1 criteria as determined by blinded independent review. Independent review is particularly relevant for a progression endpoint because assessment of tumor progression can otherwise be influenced by knowledge of treatment assignment.
RECIST version 1.1
The registry identifies RECIST version 1.1 as the tumor-assessment framework underlying the PFS definition. In this setting, the statistical endpoint depends on a clinical or radiologic assessment being translated into a documented progression event under the prespecified response-assessment rules.
9. Statistical Methods Explained
Why is PFS a time-to-event endpoint?
PFS records not only whether a participant progressed or died, but also when the first event occurred. Participants who remain event-free at the analysis cutoff can contribute partial follow-up through censoring. This makes time-to-event methods more informative than simply counting the proportion who progressed.
Why use Kaplan-Meier estimation?
Kaplan-Meier estimation is designed for settings where some participants have not yet experienced the event when observation ends. Rather than treating those participants as if they had experienced the event, their available event-free follow-up is retained up to the censoring point.
Why would a hazard ratio be useful?
A hazard ratio provides a compact relative comparison of event rates over follow-up under a Cox model. For example, an HR below 1 for the Combo 450 Group would correspond to a lower estimated instantaneous rate of progression or death than the comparator. The HR would still need its confidence interval and the context of the model assumptions to be interpreted appropriately.
Why does censoring need a prespecified rule?
Because PFS depends on whether an observation is classified as an event or censored, changing the censoring rule can change the estimated PFS distribution. The COLUMBUS registry definition explicitly references documented progression, death, initiation of new anti-cancer therapy, and censoring, making the event-definition framework central to the endpoint.
Why does blinded independent review matter?
Progression is not always a purely objective event like death. Tumor measurements and RECIST-based assessments require interpretation. A blinded independent review process is intended to reduce the possibility that knowledge of assigned treatment influences the assessment used to define progression.
Why are the two primary comparisons not automatically interchangeable?
The registry defines two separate primary comparisons: Combo 450 versus vemurafenib and Combo 450 versus LGX818. Each comparison answers a different scientific question because the comparator changes. A treatment effect relative to one comparator cannot automatically be interpreted as the same treatment effect relative to another comparator.
Why should an observed result not be inferred from the registry endpoint definition?
An endpoint definition tells us what was intended to be measured and how the event was defined. It does not establish the magnitude or direction of the observed treatment effect. Because no formal statistical analyses were posted, this page keeps the planned analytical framework separate from any numerical efficacy conclusion.
10. Analysis Populations and Interpretation
The ClinicalTrials.gov record does not provide detailed definitions of the efficacy analysis population, safety analysis population, or protocol-specific analysis sets. That limitation matters because clinical-trial estimates can differ depending on whether participants are analyzed according to randomized assignment, treatment received, or another prespecified population.
Randomized population
The trial itself is identified as randomized, with 921 enrolled participants. The ClinicalTrials.gov record does not provide a more detailed efficacy analysis-population definition.
Safety population
Serious adverse-event data are reported with explicit affected/at-risk denominators for several named treatment cohorts.
Endpoint population
The two primary endpoints are Part 1 PFS comparisons involving the Combo 450 Group, but the ClinicalTrials.gov record does not specify a separate numerical efficacy denominator.
Interpretive consequence
Formal efficacy estimates should be taken from the corresponding prespecified analysis population rather than inferred from the overall enrollment count.
11. Safety Results
The ClinicalTrials.gov record reports serious adverse events by affected participants and participants at risk for several treatment cohorts. These are the available safety results and are presented without combining cohorts that the registry identifies separately.
| Safety cohort | Serious adverse events affected / at risk |
|---|---|
| Part 2: LGX818 300 mg QD + MEK162 45 mg B | 98/257 |
| Part 1: LGX818 450 mg QD + MEK162 45 mg BI | 84/192 |
| Part 1: LGX818 300 mg | 71/192 |
| Part 2: LGX818 300 mg | 31/84 |
| Part 1: Vemurafenib 960 mg BID | 78/186 |
The affected/at-risk format is preferable to presenting the affected counts alone because the denominators differ across cohorts. A count of 98 serious adverse events among 257 participants is not directly comparable as a rate with 31 among 84 without taking the denominators into account.
The ClinicalTrials.gov record establishes the number of participants affected by serious adverse events within each named cohort and the corresponding at-risk denominator. They do not, by themselves, establish a formal statistical comparison between cohorts. In particular, no confidence intervals, p-values, risk ratios, odds ratios, or hazard ratios for serious adverse events are provided in the posted results.
12. What the Safety Denominators Tell Us
Clinical-trial safety data are inherently exposure-dependent. A safety count should therefore be read together with its denominator and the definition of the safety population.
| Concept | Application to the registry-reported COLUMBUS data |
|---|---|
| Numerator | Participants affected by a serious adverse event. |
| Denominator | Participants at risk in the named treatment cohort. |
| Part 1 vs Part 2 | The registry reports separate cohorts and denominators; they should not be pooled without an explicit registry definition. |
| Formal comparison | No comparative statistical analysis is posted on ClinicalTrials.gov for these serious adverse-event counts. |
This distinction is especially important in a multi-part study. A pooled safety percentage can be meaningful only when the underlying cohorts, exposure definitions, and analysis population are intentionally combined. The ClinicalTrials.gov record does not provide such a pooled definition.
13. Missing Data, Censoring, and Follow-Up
The registry definition provides important information about censoring even though the ClinicalTrials.gov record does not describe a full missing-data or imputation strategy. For PFS, participants without an event at the analysis cutoff or at the start of new anti-cancer therapy were subject to censoring according to the registry definition.
Censoring
Censoring allows a participant's available event-free follow-up to contribute to the analysis without treating the participant as having experienced progression.
New anti-cancer therapy
The registered time frames explicitly mention initiation of new anti-cancer therapy as part of the PFS observation framework.
Death
Death due to any cause is incorporated into the PFS definition as an event when it occurs before documented progression.
Imputation
The ClinicalTrials.gov record does not specify a missing-data imputation method, so none is attributed to the trial here.
14. Multiplicity and the Two Primary Comparisons
COLUMBUS registered two primary PFS endpoints. Both concern the same Combo 450 Group but use different comparators. From a statistical-design perspective, multiple confirmatory comparisons raise the question of how the overall type I error rate is controlled.
The ClinicalTrials.gov record does not state an alpha allocation, multiplicity procedure, hierarchical testing strategy, or other formal error-control method. It would therefore be inappropriate to infer one from the existence of two primary endpoints.
| Primary comparison | Scientific comparison | Multiplicity information reported? |
|---|---|---|
| Primary PFS endpoint 1 | Combo 450 Group vs Vemurafenib Group | No formal procedure reported |
| Primary PFS endpoint 2 | Combo 450 Group vs LGX818 Group | No formal procedure reported |
15. Interim Analysis and Bayesian Methods
The ClinicalTrials.gov record does not identify an interim analysis plan, alpha-spending procedure, stopping boundary, or Bayesian analysis. These design features therefore cannot be attributed to COLUMBUS from the provided registry information.
Interim analysis
No interim-analysis method is specified in the ClinicalTrials.gov record.
Alpha spending
No alpha-spending or group-sequential method is specified in the ClinicalTrials.gov record.
Bayesian methods
No Bayesian statistical method is identified in the ClinicalTrials.gov record.
Interpretation
These omissions are different from saying that such methods were not used in the complete trial documentation; they simply are not reported in the ClinicalTrials.gov record used for this page.
16. Non-Inferiority, Crossover, and Factorial Design
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Non-inferiority margin | No non-inferiority margin is provided. The registered primary endpoints are PFS comparisons, but the ClinicalTrials.gov record does not characterize them as non-inferiority hypotheses. |
| Crossover | The design model is listed as parallel. No crossover procedure is provided. |
| Factorial design | The design model is listed as parallel, not factorial. |
| Randomization | The allocation is explicitly randomized. |
This distinction is useful because the statistical logic differs substantially across trial designs. A non-inferiority trial requires a prespecified margin and a one-sided inferential framework; a crossover trial requires within-participant treatment-period considerations; and a factorial trial generally evaluates more than one intervention factor. None of those additional design structures is established by the registry-reported COLUMBUS record.
17. Trial History
Trial start
The registry lists 2013-09-12 as the study start date.
Primary completion
The registry lists 2016-11-09 as the primary completion date.
Registry status
The trial status is listed as COMPLETED, with results posted and 37 outcome measures posted.
The ClinicalTrials.gov record does not provide publication dates, individual analysis cutoffs, or a sequence of efficacy-result releases. The five linked PubMed records are therefore included as source links without assigning them to specific analyses or dates not contained in the trial data.
18. Why This Trial Matters Statistically
COLUMBUS is a useful teaching case because its design combines randomization, multiple parallel treatment groups, blinded independent assessment of a time-to-event endpoint, and more than one primary comparison. Even without importing numerical efficacy estimates from external publications, the registry structure illustrates several important principles in clinical-trial statistics.
| Statistical concept | How it appears in COLUMBUS |
|---|---|
| Randomization | The trial allocation is explicitly randomized. |
| Parallel-group design | The design model is explicitly parallel. |
| Time-to-event analysis | Both registered primary endpoints are PFS endpoints. |
| Kaplan-Meier estimation | A standard approach for estimating PFS distributions with censoring. |
| Hazard ratio | A standard effect measure for comparing time-to-event rates under a Cox model. |
| Log-rank testing | A standard method for comparing survival distributions between randomized groups. |
| Independent review | Primary PFS endpoints are defined by blinded independent review. |
| RECIST version 1.1 | The registry identifies RECIST version 1.1 for tumor assessment. |
| Multiple primary comparisons | Two primary PFS comparisons are registered. |
| Safety denominators | Serious adverse events are reported as affected/at-risk counts across named cohorts. |
19. Important Limitations and Interpretation Issues
- No formal statistical analyses reported: no formal statistical analyses were posted, so comparative efficacy estimates cannot be reported from the ClinicalTrials.gov record.
- Posted results are not the same as posted comparative estimates: the registry indicates that results are posted, but the ClinicalTrials.gov recordset does not include formal effect estimates for the primary PFS endpoints.
- Multiple primary comparisons: two primary PFS comparisons are registered, but the ClinicalTrials.gov record does not specify how multiplicity was handled.
- Analysis population: the ClinicalTrials.gov record does not provide a detailed efficacy analysis-population definition.
- Censoring: PFS interpretation depends on the registry's event and censoring rules, including the treatment of initiation of new anti-cancer therapy.
- Proportional-hazards assumption: if a Cox model is used, its hazard ratio interpretation depends on the model's assumptions. The ClinicalTrials.gov record does not report a proportional-hazards assessment.
- Safety cohort structure: serious adverse events are reported for separately named Part 1 and Part 2 cohorts, so the denominators should not be pooled without explicit justification.
- External publication results: this page does not import numerical efficacy results from the linked PubMed records because the task specifies that trial numbers come only from the ClinicalTrials.gov record.
20. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The trial has two registered randomized PFS comparisons. Both are time-to-event endpoints assessed by blinded independent review, and their analysis would ordinarily require methods that accommodate censoring.
Clinical interpretation
The ClinicalTrials.gov record establishes the treatment groups, endpoint definitions, study structure, and safety counts. They do not provide enough numerical comparative efficacy information to quantify a treatment effect.
This distinction prevents a common analytical error: treating a well-defined endpoint and a completed randomized trial as evidence of a particular numerical treatment effect when the actual effect estimate has not been reported. A rigorous statistical review should report what is known, identify what is missing, and avoid filling gaps with external assumptions.
21. What Would Be Needed for a Complete Numerical PFS Analysis?
A complete quantitative analysis of the two primary PFS endpoints would require the formal statistical results associated with each comparison. At minimum, the analysis would normally need an effect estimate, its confidence interval, and the inferential result under the prespecified testing framework.
| Information | Why it matters |
|---|---|
| Hazard ratio | Quantifies the estimated relative event rate between treatment groups under a Cox model. |
| 95% confidence interval | Describes statistical uncertainty around the estimated hazard ratio. |
| P-value | Provides the inferential result for the specified null hypothesis under the chosen testing framework. |
| Median PFS | Provides an absolute time-scale summary of the PFS distribution when estimable. |
| Kaplan-Meier estimates | Describe PFS probability over selected follow-up times. |
| Number of events | Provides information about the amount of event information contributing to the analysis. |
| Analysis population | Identifies which participants contributed to the formal efficacy analysis. |
| Multiplicity procedure | Determines how the two primary comparisons are interpreted within the overall type I error framework. |
22. Related Tutorials
Learn more about the statistical methods and concepts used to understand this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: COLUMBUS, NCT01909453.
- Linked PubMed record: PubMed 37534686.
- Linked PubMed record: PubMed 37506329.
- Linked PubMed record: PubMed 37309702.
- Linked PubMed record: PubMed 36653848.
- Linked PubMed record: PubMed 35862871.
Continue learning through Clinical Biostats
Explore statistical tutorials and calculators for randomized trials, survival analysis, time-to-event endpoints, and related methods.
25. Record Summary
COLUMBUS is a completed randomized phase 3 parallel-group trial in melanoma with 921 enrolled participants and four arms. Its two registered primary endpoints are time-to-event PFS comparisons by blinded independent review, with the Combo 450 Group compared separately with the vemurafenib group and the LGX818 group. The registry defines PFS around documented disease progression or death and identifies RECIST version 1.1, independent review, censoring, and initiation of new anti-cancer therapy within the endpoint framework.
The ClinicalTrials.gov record also provide serious adverse-event counts for several named Part 1 and Part 2 treatment cohorts. Those safety results are presented using their reported affected/at-risk denominators rather than being pooled across cohorts.