This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the information contained in the posted registry data summarized for ALTA-1L.
1. Trial at a Glance
ALTA-1L was a randomized, parallel, open-label phase 3 trial comparing brigatinib with crizotinib in participants with ALK-positive advanced NSCLC. The registry reports 275 participants enrolled and a primary endpoint of progression-free survival assessed by a Blinded Independent Review Committee.
| Feature | ALTA-1L |
|---|---|
| Trial name | ALTA-1L |
| Brief title | ALTA-1L Study: A Study of Brigatinib Versus Crizotinib in Anaplastic Lymphoma Kinase Positive (ALK+) Advanced Non-small Cell Lung Cancer (NSCLC) Participants |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Oncology |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 275.0 |
| Interventions | Brigatinib; Crizotinib |
| Primary endpoint | Progression-free Survival (PFS) |
| Primary endpoint type | Time-to-event |
| Hypothesis type | Superiority |
| Lead sponsor | Ariad Pharmaceuticals |
| Sponsor type | Industry |
| Trial start | 2016-05-26 |
| Primary completion | 2020-07-28 |
| ClinicalTrials.gov | NCT02737501 |
2. Clinical Question
The central randomized question was whether brigatinib produced a different progression-free survival experience from crizotinib in participants with ALK-positive advanced NSCLC. The registered hypothesis type was superiority.
Population
Participants with anaplastic lymphoma kinase positive (ALK+) advanced non-small cell lung cancer (NSCLC).
Intervention
Brigatinib, with the randomized-phase regimen represented in the statistical analysis as brigatinib 90 mg QD/180 QD.
Comparator
Crizotinib 250 mg BID in the randomized phase.
Primary question
Does brigatinib improve progression-free survival relative to crizotinib under the prespecified superiority comparison?
3. Trial Design
The registry describes ALTA-1L as a randomized, parallel, unmasked phase 3 treatment trial. Three arms are represented in the ClinicalTrials.gov record, while the posted statistical comparisons principally contrast the randomized brigatinib and randomized crizotinib groups. A separate crossover phase is also represented in the safety data.
Brigatinib
- Brigatinib 90 mg QD/180 QD
- Compared with crizotinib in the randomized efficacy analyses
- Primary efficacy analysis conducted in the ITT population
Crizotinib
- Crizotinib 250 mg BID
- Randomized comparator for the primary PFS analysis
- Primary efficacy analysis conducted in the ITT population
4. Randomization, Stratification, and Analysis Population
The posted primary analysis used the intention-to-treat population. The registry defines this population as all participants randomized to each regimen regardless of whether they tested ALK+, whether they received study drug, or whether they adhered to the assigned dose.
| Feature | Registry-supported description |
|---|---|
| Analysis population | ITT population |
| ITT definition | All participants randomized to each regimen regardless of whether they tested ALK+, received study drug, or adhered to the assigned dose |
| Primary comparison | Randomized-phase brigatinib 90 mg QD/180 QD versus randomized-phase crizotinib 250 mg BID |
| Primary endpoint analysis | Log-rank test |
| Hazard-ratio model | Cox proportional-hazards model |
| Covariate adjustment | Randomization stratification factors in the primary PFS Cox model |
The registry analysis notes that the PFS hazard ratio was obtained from a Cox proportional-hazards model with the current randomization stratification factors as covariates. The ClinicalTrials.gov record does not identify those factors individually, so they are not specified here.
5. Endpoints
| Endpoint | Registry definition / time frame | Type |
|---|---|---|
| Progression-free Survival (PFS) | PFS as assessed by Blinded Independent Review Committee (BIRC), per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1, was defined as the time interval from the date of randomization until the date of the first documented PD event. The data was censored for participants without a PFS event. Time frame: Up to end of study (Up to 56 months). | Time-to-event |
| Confirmed Objective Response Rate (ORR) | Baseline up to end of treatment (Up to 36 months) | Binary |
| Confirmed Intracranial ORR (iORR) | Baseline up to end of treatment (Up to 36 months) | Binary |
| Intracranial Progression Free Survival | Baseline up to end of study (Up to 56 months) | Time-to-event |
| Disease Control Rate (DCR) | Baseline up to end of treatment (Up to 36 months) | Binary |
| Change From Baseline in Global Health Status/Quality of Life as Assessed by EORTC QLQ-C30 (Version 3.0) | Baseline and Month 36 | Continuous |
The registry profile reports 1 primary endpoint, 10 posted outcome measures, and 6 posted statistical analyses. The primary endpoint is the time-to-event PFS measure defined above.
6. Statistical Methodology
Log-rank test for progression-free survival
The primary PFS comparison used a log-rank test. This is a standard method for comparing time-to-event distributions between randomized groups while retaining information from participants who are censored before experiencing the event.
Conceptually, the log-rank test compares the observed number of events in each group with the number expected under a null hypothesis of equivalent survival distributions. Its natural setting is a randomized comparison in which follow-up times may differ and some participants do not experience the event during observation.
For ALTA-1L, the registered PFS definition specifically uses time from randomization to first documented PD, with censoring for participants without a PFS event.
Cox proportional-hazards model
The reported PFS hazard ratio was obtained from a Cox proportional-hazards model with the current randomization stratification factors included as covariates. This provides a model-based relative comparison of the instantaneous event rates between the randomized regimens.
Equivalently, 1 − 0.481 = 0.519, so the estimate corresponds to an approximately 51.9% lower estimated hazard under the fitted model. This is a relative hazard interpretation, not a statement that 51.9% of participants avoided progression.
Cochran-Mantel-Haenszel test
The confirmed ORR, confirmed intracranial ORR, and DCR analyses used the Cochran-Mantel-Haenszel (CMH) test. This approach allows a categorical treatment comparison while accounting for specified strata rather than simply pooling all participants into an unstratified 2×2 analysis.
Mixed-effects model
Change from baseline in global health status/quality of life was analyzed with a mixed-effects model. The registry specifies an unstructured covariance matrix. Mixed-effects models are useful for longitudinal outcomes because observations from the same participant are correlated rather than statistically independent.
The unstructured covariance specification allows the model to represent the covariance pattern among repeated observations without imposing a simple equal-correlation structure.
Intention-to-treat analysis
The posted analyses consistently identify the ITT population. The practical statistical principle is important: once treatment is randomized, analyzing participants according to their assigned regimen preserves the treatment comparison generated by randomization, even when actual treatment exposure or adherence differs.
7. Primary Result: Progression-Free Survival
The primary endpoint was PFS assessed by BIRC according to RECIST version 1.1, with follow-up specified as up to the end of study, up to 56 months. The analysis compared randomized-phase brigatinib 90 mg QD/180 QD with randomized-phase crizotinib 250 mg BID in the ITT population.
Hazard ratio for progression-free survival
95% CI: 0.35–0.66 · P < 0.0001
Two-sided confidence interval · Superiority hypothesis
| Primary endpoint | Brigatinib vs crizotinib |
|---|---|
| Endpoint | Progression-free Survival (PFS) |
| Time frame | Up to end of study (Up to 56 months) |
| Analysis population | ITT population |
| Method | Log-rank test |
| Effect measure | Hazard ratio |
| Estimate | 0.481 |
| 95% CI | 0.35–0.66 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
| Model note | Cox proportional-hazards model with randomization stratification factors as covariates |
The estimated PFS hazard ratio of 0.481 indicates that, under the fitted Cox model, the estimated instantaneous rate of the PFS event was about 48.1% as high in the brigatinib group as in the crizotinib group. Expressed as a relative hazard reduction, the estimate corresponds to approximately 51.9%.
This does not mean that 51.9% of participants were protected from progression, that every participant experienced the same reduction, or that median PFS was reduced or increased by 51.9%. A hazard ratio summarizes a relative event-rate comparison over the modeled follow-up; it is not an absolute risk difference.
The 95% confidence interval of 0.35–0.66 describes statistical uncertainty around the estimated hazard ratio under the model and sampling framework. It does not describe the range of individual treatment effects across patients.
The P-value < 0.0001 addresses evidence against the relevant null hypothesis under the statistical test. It does not measure the size or clinical importance of the effect. Effect magnitude is better conveyed by the hazard ratio itself together with its confidence interval and, where available, absolute survival estimates.
Because the hazard ratio comes from a Cox proportional-hazards model, interpretation also depends on the proportional-hazards framework. The ClinicalTrials.gov record does not provide a time-varying hazard-ratio assessment, so the single HR should not be interpreted as proving that the relative event-rate ratio was constant at every point in follow-up.
8. Secondary Result: Confirmed Objective Response Rate
Confirmed ORR was a secondary binary endpoint measured from baseline up to the end of treatment, up to 36 months. The analysis compared the same randomized-phase brigatinib and crizotinib groups in the ITT population.
Odds ratio for confirmed ORR
95% CI: 1.04–2.91 · P = 0.0330
Two-sided confidence interval · Superiority hypothesis
| Feature | Posted analysis |
|---|---|
| Endpoint | Confirmed Objective Response Rate (ORR) |
| Time frame | Baseline up to end of treatment (Up to 36 months) |
| Endpoint type | Binary |
| Analysis population | ITT population |
| Method | Cochran-Mantel-Haenszel test |
| Effect measure | Odds ratio |
| Estimate | 1.74 |
| 95% CI | 1.04–2.91 |
| P-value | 0.0330 |
| Stratification | Presence of iCNS metastases at Baseline, and prior chemotherapy for locally advanced or metastatic disease |
An odds ratio of 1.74 means that the estimated odds of confirmed objective response were 1.74 times as high in the brigatinib group as in the crizotinib group under the stratified CMH analysis.
An odds ratio is not the same as a risk ratio or percentage-point difference. Without the underlying response proportions, an OR of 1.74 should not be translated into a particular percentage increase in response probability.
The 95% CI of 1.04–2.91 indicates uncertainty around the estimated odds ratio. Its lower bound is close to 1, so the numerical estimate should not be read as highly precise despite the reported P-value of 0.0330.
The P-value describes evidence against the null comparison under the CMH testing framework. It does not tell us that the effect is "3.3%" in size, nor does it describe the probability that brigatinib is effective.
The analysis was stratified by baseline iCNS metastases and prior chemotherapy. That matters because a stratified CMH analysis compares the treatment groups while accounting for the distribution of participants across those specified strata rather than treating the entire population as a single homogeneous 2×2 table.
9. Secondary Result: Confirmed Intracranial ORR
Confirmed intracranial ORR was analyzed as a binary secondary endpoint from baseline up to the end of treatment, up to 36 months. The registry reports a CMH analysis and an odds ratio of 13.56.
Odds ratio for confirmed intracranial ORR
95% CI: 4.70–39.11 · P < 0.0001
Two-sided confidence interval · Superiority hypothesis
| Feature | Posted analysis |
|---|---|
| Endpoint | Confirmed Intracranial ORR (iORR) |
| Time frame | Baseline up to end of treatment (Up to 36 months) |
| Endpoint type | Binary |
| Analysis population | ITT population |
| Method | Cochran-Mantel-Haenszel test |
| Effect measure | Odds ratio |
| Estimate | 13.56 |
| 95% CI | 4.70–39.11 |
| P-value | <0.0001 |
| Stratification | Presence of prior chemotherapy for locally advanced or metastatic disease at study entry |
The odds ratio of 13.56 indicates substantially higher estimated odds of confirmed intracranial response in the brigatinib group relative to the crizotinib group under the posted stratified CMH analysis.
This is an odds comparison, not a claim that intracranial response was 13.56 times as common in terms of probability. The relationship between odds and probability depends on the underlying event rate.
The 95% CI of 4.70–39.11 is wide. It indicates that the point estimate is not highly precise even though the entire reported interval is above 1. The width of the interval is an important part of interpreting an effect of this magnitude.
The P-value of <0.0001 indicates strong evidence against the null hypothesis in the posted statistical test. It does not eliminate uncertainty about the magnitude of the effect, which is why the confidence interval remains essential.
The analysis was stratified by prior chemotherapy for locally advanced or metastatic disease at study entry. As with the ORR analysis, the stratification is part of the statistical design and should be retained when explaining what the reported odds ratio represents.
10. Secondary Result: Intracranial Progression-Free Survival
Intracranial progression-free survival was a secondary time-to-event endpoint evaluated from baseline up to the end of study, up to 56 months. The registry reports a log-rank analysis and a Cox-model hazard ratio adjusted for prior chemotherapy for locally advanced or metastatic disease.
Hazard ratio for intracranial progression-free survival
95% CI: 0.17–0.51 · P < 0.0001
Two-sided confidence interval · Superiority hypothesis
| Feature | Posted analysis |
|---|---|
| Endpoint | Intracranial Progression Free Survival |
| Time frame | Baseline up to end of study (Up to 56 months) |
| Endpoint type | Time-to-event |
| Analysis population | ITT population |
| Method | Log-rank test |
| Effect measure | Hazard ratio |
| Estimate | 0.293 |
| 95% CI | 0.17–0.51 |
| P-value | <0.0001 |
| Model note | Cox proportional-hazards model with prior chemotherapy for locally advanced or metastatic disease as covariate |
The hazard ratio of 0.293 corresponds to an estimated intracranial event hazard about 29.3% as high in the brigatinib group as in the crizotinib group under the fitted model. Equivalently, 1 − 0.293 = 0.707, corresponding to an approximately 70.7% lower estimated hazard.
This does not mean that 70.7% of intracranial events were prevented, nor does it provide an absolute probability of intracranial progression. The hazard ratio is a relative time-to-event measure.
The 95% CI of 0.17–0.51 indicates uncertainty around the estimate. The interval remains well below 1, while its width shows that the exact magnitude of the relative effect is not known with arbitrary precision.
The P-value of <0.0001 is evidence against the null hypothesis in the reported log-rank framework. It is not a measure of the effect size and should not be used as a substitute for the hazard ratio or confidence interval.
The same Cox-model cautions apply here as to the primary PFS analysis: the HR is model-based and its interpretation should not be extended to individual patients or treated as proof of a constant relative hazard throughout follow-up.
11. Secondary Result: Disease Control Rate
Disease control rate was a binary secondary endpoint measured from baseline up to the end of treatment, up to 36 months. The posted analysis used a stratified Cochran-Mantel-Haenszel test.
Odds ratio for disease control
95% CI: 0.47–1.82 · P = 0.8220
Two-sided confidence interval · Superiority hypothesis
| Feature | Posted analysis |
|---|---|
| Endpoint | Disease Control Rate (DCR) |
| Time frame | Baseline up to end of treatment (Up to 36 months) |
| Endpoint type | Binary |
| Analysis population | ITT population |
| Method | Cochran-Mantel-Haenszel test |
| Effect measure | Odds ratio |
| Estimate | 0.93 |
| 95% CI | 0.47–1.82 |
| P-value | 0.8220 |
| Stratification | Presence of iCNS metastases at Baseline, and prior chemotherapy for locally advanced or metastatic disease |
The DCR odds ratio of 0.93 is close to 1, indicating that the estimated odds of disease control were similar between groups under the posted stratified analysis.
The 95% CI of 0.47–1.82 spans 1 and is relatively broad. Thus, the estimate is compatible with a range of relative odds in either direction under the statistical uncertainty represented by the interval.
The reported P-value of 0.8220 does not provide evidence against the null hypothesis in the posted CMH test. It should not be interpreted as proof that the two treatments are identical; failure to reject a null hypothesis is not the same as establishing equivalence.
The contrast between this DCR result and the posted response and time-to-event results is also statistically instructive. Different endpoints capture different aspects of disease behavior. A treatment comparison can differ across response definitions and time-to-event endpoints because they do not measure the same outcome.
12. Secondary Result: Global Health Status / Quality of Life
Change from baseline in global health status/quality of life, assessed with EORTC QLQ-C30 Version 3.0, was evaluated at baseline and Month 36. The registry reports a mixed-effects model with an unstructured covariance matrix.
Least square mean difference
95% CI: 0.58–11.00 · P = 0.0295
Two-sided confidence interval · Superiority hypothesis
| Feature | Posted analysis |
|---|---|
| Endpoint | Change From Baseline in Global Health Status/Quality of Life as Assessed by EORTC QLQ-C30 (Version 3.0) |
| Time frame | Baseline and Month 36 |
| Endpoint type | Continuous |
| Analysis population | ITT population |
| Method | Mixed Models Analysis |
| Effect measure | Least Square Mean Difference |
| Estimate | 5.79 |
| 95% CI | 0.58–11.00 |
| P-value | 0.0295 |
| Model specification | Mixed-effects model with an unstructured covariance matrix |
The reported least square mean difference of 5.79 is a model-based adjusted difference between randomized groups for the specified quality-of-life outcome. It is not the same statistical quantity as an odds ratio or hazard ratio.
The 95% CI of 0.58–11.00 expresses uncertainty around that estimated mean difference. Because the registry describes the endpoint as a score on a scale, the interpretation of the numerical magnitude depends on the properties and scoring conventions of the EORTC QLQ-C30 measure.
The P-value of 0.0295 indicates evidence against the null hypothesis in the posted mixed-model analysis. It does not quantify the clinical importance of a 5.79-point model-based difference.
The mixed-effects framework is particularly relevant because repeated measurements from the same participant are correlated. The unstructured covariance matrix allows the analysis to model that within-participant correlation without forcing a simpler covariance pattern.
13. Results Summary
| Endpoint | Effect measure | Estimate | 95% CI | P-value | Method |
|---|---|---|---|---|---|
| Progression-free Survival (PFS) | Hazard ratio | 0.481 | 0.35–0.66 | <0.0001 | Log-rank test; Cox model |
| Confirmed Objective Response Rate (ORR) | Odds ratio | 1.74 | 1.04–2.91 | 0.0330 | Cochran-Mantel-Haenszel |
| Confirmed Intracranial ORR (iORR) | Odds ratio | 13.56 | 4.70–39.11 | <0.0001 | Cochran-Mantel-Haenszel |
| Intracranial Progression Free Survival | Hazard ratio | 0.293 | 0.17–0.51 | <0.0001 | Log-rank test; Cox model |
| Disease Control Rate (DCR) | Odds ratio | 0.93 | 0.47–1.82 | 0.8220 | Cochran-Mantel-Haenszel |
| Global Health Status / Quality of Life | Least square mean difference | 5.79 | 0.58–11.00 | 0.0295 | Mixed-effects model |
The table illustrates why a clinical-trial statistical summary should not collapse every endpoint into a single "positive" or "negative" label. The endpoints use different estimands and statistical models. The primary PFS result is a time-to-event comparison, response and disease control are odds-based categorical comparisons, intracranial PFS is another time-to-event analysis, and quality of life is a longitudinal continuous outcome.
14. Statistical Methods Explained
Why was a log-rank test used for PFS?
PFS is a time-to-event endpoint, and participants can have different follow-up times or be censored without experiencing a documented event. The log-rank test is designed for comparing the resulting survival distributions while accounting for the timing of events and censoring. It therefore matches the structure of the registered PFS endpoint much better than a simple comparison of proportions.
What does a PFS hazard ratio of 0.481 mean?
It means the fitted model estimates the instantaneous PFS event hazard in the brigatinib group at about 48.1% of the crizotinib-group hazard. The complementary expression, 51.9%, is an estimated relative hazard reduction. It is not a 51.9-percentage-point improvement in PFS probability and is not a statement about the proportion of individual patients who benefited.
Why was a Cochran-Mantel-Haenszel test used for ORR?
ORR is binary, so each participant contributes a response classification rather than a survival time. The CMH framework allows the treatment comparison to account for prespecified strata. In ALTA-1L, the posted ORR analysis was stratified by baseline iCNS metastases and prior chemotherapy for locally advanced or metastatic disease.
Why can an odds ratio be very different from a risk ratio?
Odds and probability are different mathematical quantities. If response is uncommon, an odds ratio and risk ratio can be numerically similar; as the outcome becomes more common, they can diverge substantially. Therefore an OR of 13.56 for intracranial response should not be described as meaning that the probability of response was 13.56 times higher without knowing the underlying probabilities.
Why was a mixed-effects model used for quality of life?
Quality-of-life outcomes can be measured repeatedly in the same participant. Those observations are correlated, so treating them as independent would generally misrepresent the variance structure. A mixed-effects model provides a framework for modeling treatment effects while accounting for within-participant correlation. The registry specifically reports an unstructured covariance matrix.
Why does the ITT population matter?
The ITT principle retains participants in their randomized groups for efficacy analysis. This preserves the treatment comparison created by randomization rather than allowing subsequent adherence or treatment exposure to redefine the comparison. In ALTA-1L, the registry explicitly defines the ITT population as all randomized participants regardless of whether they tested ALK+, received study drug, or adhered to the assigned dose.
Why is a confidence interval needed when the P-value is very small?
A P-value answers an evidence-against-the-null question; it does not describe how large the treatment effect is or how precisely it has been estimated. For example, the PFS P-value is <0.0001, but the 95% CI of 0.35–0.66 is still necessary to communicate the range of model-based uncertainty around the HR estimate.
15. Understanding the Primary PFS Analysis
What was compared?
Randomized-phase brigatinib 90 mg QD/180 QD versus randomized-phase crizotinib 250 mg BID.
Who was analyzed?
The ITT population, including all randomized participants under the registry's definition.
What was measured?
Time from randomization to the first documented PD event, with censoring for participants without a PFS event.
How was it tested?
Log-rank test, with the HR obtained from a Cox proportional-hazards model using randomization stratification factors as covariates.
This sequence is important because the numerical HR cannot be interpreted independently of the estimand and analysis population. The same number would have a different meaning if it represented an odds ratio, a risk ratio, or an unadjusted event proportion.
16. Odds Ratios vs Hazard Ratios in ALTA-1L
| Measure | Endpoints | What it describes |
|---|---|---|
| Hazard ratio | PFS; intracranial PFS | Relative comparison of instantaneous event hazards in a time-to-event model |
| Odds ratio | Confirmed ORR; confirmed iORR; DCR | Relative comparison of odds for binary outcomes |
| Least square mean difference | Global health status / quality of life | Model-based difference in the estimated mean outcome between randomized groups |
These effect measures should not be interchanged. A hazard ratio incorporates event timing and censoring; an odds ratio describes a binary outcome; and a least square mean difference describes a modeled continuous outcome. A statistically rigorous trial summary therefore keeps the estimand matched to the endpoint.
17. Multiplicity and Interpretation Across Endpoints
The registry reports one primary endpoint and multiple secondary statistical analyses. That structure creates an important distinction between the primary confirmatory question and the collection of additional endpoint comparisons.
| Endpoint role | Endpoint | Interpretive issue |
|---|---|---|
| Primary | Progression-free Survival (PFS) | Primary superiority comparison |
| Secondary | Confirmed ORR | Binary response comparison |
| Secondary | Confirmed iORR | Intracranial binary response comparison |
| Secondary | Intracranial Progression Free Survival | Time-to-event comparison |
| Secondary | Disease Control Rate | Binary disease-control comparison |
| Secondary | Global Health Status / Quality of Life | Longitudinal continuous outcome |
The posted data identify the hypothesis type for these analyses as superiority. The ClinicalTrials.gov record does not provide a detailed multiplicity-adjustment scheme across all six posted analyses. Consequently, the individual P-values should be interpreted as the reported results of their respective analyses rather than assuming an unreported familywise-error procedure.
18. Safety Results
The ClinicalTrials.gov record reports serious adverse events by arm as affected participants among participants at risk. These data include the randomized brigatinib group, randomized crizotinib group, and crossover-phase brigatinib group.
| Phase / arm | Serious adverse events affected | At risk |
|---|---|---|
| Randomized Phase: Brigatinib 90 mg QD/180 | 56 | 136 |
| Randomized Phase: Crizotinib 250 mg BID | 53 | 137 |
| Crossover Phase: Brigatinib 90 mg QD | 24 | 65 |
The serious-adverse-event data are reported separately from the efficacy analyses. That distinction is important because the primary efficacy comparison is based on randomization and the ITT population, whereas safety summaries can depend on treatment exposure and the population defined for the particular safety measure.
Randomized brigatinib
56 affected participants among 136 at risk for the reported serious-adverse-event measure.
Randomized crizotinib
53 affected participants among 137 at risk for the reported serious-adverse-event measure.
Crossover brigatinib
24 affected participants among 65 at risk for the reported serious-adverse-event measure.
Interpretive caution
The crossover-phase group should not be interpreted as an independently randomized comparator arm.
19. Why the ITT Definition Is Especially Important Here
The registry's ITT definition is unusually explicit: participants remain in their randomized treatment groups regardless of whether they tested ALK+, received study drug, or adhered to the assigned dose.
For a randomized efficacy comparison, this is statistically important because randomization is what creates the basis for a causal comparison. Removing participants after randomization based on post-randomization treatment exposure can disturb that balance and introduce selection into the analysis.
At the same time, ITT does not make every analysis problem disappear. Treatment discontinuation, adherence, crossover, missing assessments, and subsequent therapies can all affect how an endpoint should be interpreted. The ITT principle preserves the randomized comparison, but it does not eliminate the need to understand what happened after randomization.
20. Time-to-Event Endpoints and Censoring
The registered PFS definition states that participants are followed from randomization until the first documented PD event, with data censored for participants without a PFS event. This is a defining feature of time-to-event analysis.
A censored participant is not treated as if the event occurred at the censoring date. Instead, the participant contributes information about being event-free up to that point.
Kaplan-Meier estimation is a common descriptive tool for this type of data, while the log-rank test provides a group comparison and the Cox model provides a hazard-ratio framework. The registry's posted PFS analysis specifically reports the log-rank test and Cox proportional-hazards model.
21. Statistical Interpretation of the P-values
ALTA-1L illustrates why a P-value should be read together with an effect estimate and its confidence interval.
| Endpoint | Estimate | P-value | What the P-value does not tell you |
|---|---|---|---|
| PFS | HR 0.481 | <0.0001 | It does not tell you the magnitude of the HR or an absolute PFS probability. |
| Confirmed ORR | OR 1.74 | 0.0330 | It does not equal a 3.3% treatment effect or give the response probability. |
| Confirmed iORR | OR 13.56 | <0.0001 | It does not mean a 13.56-fold increase in response probability. |
| Intracranial PFS | HR 0.293 | <0.0001 | It does not state the proportion of participants who avoided intracranial progression. |
| DCR | OR 0.93 | 0.8220 | It does not prove equivalence or identical treatment effects. |
| Quality of life | LS mean difference 5.79 | 0.0295 | It does not by itself establish clinical importance. |
A useful statistical reading therefore follows the sequence endpoint → estimand → estimate → confidence interval → P-value → analysis population → model assumptions. This is more informative than sorting results according to P-value alone.
22. Limitations
- Registry-level detail: the ClinicalTrials.gov record provides the main posted analyses but do not include every element of a full statistical analysis plan.
- No median PFS: the ClinicalTrials.gov record does not report median PFS, so no median survival estimate is presented here.
- No underlying survival data: the ClinicalTrials.gov record does not contain participant-level event and censoring times, preventing independent reconstruction of a Kaplan-Meier curve.
- Hazard-ratio assumptions: the Cox proportional-hazards model is a model-based analysis. The ClinicalTrials.gov record does not provide a formal assessment of the proportional-hazards assumption.
- Multiple endpoints: six statistical analyses are posted, but the ClinicalTrials.gov record does not specify a complete multiplicity-adjustment hierarchy for all of them.
- Analysis populations: the efficacy analyses use ITT, while the serious-adverse-event data are reported by affected and at-risk participants for specific trial phases.
- Crossover: a crossover-phase brigatinib group is present in the safety data. This phase should not be interpreted as a separately randomized comparison.
- Odds-ratio interpretation: ORs cannot be translated directly into probability ratios without the underlying event probabilities.
- Quality-of-life interpretation: the statistical difference is a least square mean difference from a mixed-effects model; the ClinicalTrials.gov record does not provide additional information needed to establish a clinical-importance threshold.
- Stratification detail: the PFS model is described as using current randomization stratification factors, but the registry summary does not identify all of those factors individually.
23. Why This Trial Matters Statistically
ALTA-1L is a useful statistical teaching example because one randomized clinical trial contains several distinct analytical structures. Its primary endpoint is a time-to-event outcome analyzed with a log-rank test and Cox model, while several secondary endpoints use stratified categorical analysis and another uses a longitudinal mixed-effects model.
| Statistical concept | How it appears in ALTA-1L |
|---|---|
| Randomization | Randomized phase 3 parallel-group comparison |
| Intention-to-treat analysis | Primary and posted secondary efficacy analyses use the ITT population |
| Time-to-event analysis | PFS and intracranial PFS |
| Log-rank test | Primary PFS and intracranial PFS analyses |
| Hazard ratio | Effect measure for PFS and intracranial PFS |
| Cox model | Hazard-ratio estimation with covariate adjustment |
| Stratified analysis | Used in PFS modeling and CMH analyses |
| Cochran-Mantel-Haenszel test | ORR, iORR, and DCR |
| Odds ratio | Binary response and disease-control endpoints |
| Mixed-effects model | Global health status / quality-of-life outcome |
| Unstructured covariance | Specified for the mixed-effects quality-of-life analysis |
| Confidence intervals | Reported for every posted statistical effect estimate |
| Multiple endpoints | One primary endpoint plus multiple secondary analyses |
| Crossover phase | Separate crossover-phase brigatinib group in the safety data |
24. What the Effect Estimates Say Together
The six posted statistical analyses do not all point in exactly the same numerical direction, and that is itself informative. The primary PFS HR is 0.481, intracranial PFS HR is 0.293, ORR OR is 1.74, and iORR OR is 13.56. DCR has an OR of 0.93, while the quality-of-life analysis reports a least square mean difference of 5.79.
These estimates should not be averaged, ranked, or combined into a single overall statistic. They correspond to different endpoints, different effect measures, and different statistical models. A hazard ratio of 0.481 and an odds ratio of 1.74 do not live on the same numerical scale and cannot be compared as if they were interchangeable measures.
The most statistically coherent summary is therefore endpoint-specific: the primary PFS analysis estimates a lower relative event hazard with brigatinib; the intracranial PFS analysis estimates an even lower relative intracranial event hazard; the response analyses estimate treatment-group differences in odds of response; DCR produces an odds estimate close to the null; and quality of life is represented by a model-based mean difference.
25. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The randomized PFS comparison produced a hazard ratio of 0.481 with a 95% CI of 0.35–0.66 and P < 0.0001. Secondary analyses used appropriate endpoint-specific models and reported corresponding estimates and confidence intervals.
Endpoint-specific interpretation
The trial also reports odds-based response measures, an intracranial time-to-event measure, a disease-control measure, and a longitudinal quality-of-life outcome. Each requires interpretation according to its own estimand.
Absolute effects
The ClinicalTrials.gov record does not report median PFS or time-specific survival probabilities, so this page does not infer absolute PFS values from the hazard ratio.
Safety interpretation
Serious adverse events are reported separately by randomized and crossover phases. Safety findings should be interpreted independently from efficacy effect estimates.
26. Longitudinal Trial History
Trial start
The ALTA-1L trial began on May 26, 2016.
Randomized comparison
The trial used randomized, parallel allocation with no masking and compared brigatinib with crizotinib.
Primary completion
The registry lists July 28, 2020 as the primary completion date.
Results posted
The registry profile identifies the trial as completed and reports six statistical analyses, including the primary PFS analysis and five secondary analyses summarized on this page.
27. A Deeper Look at the Primary Hazard Ratio
The PFS HR of 0.481 is a ratio of modeled event hazards. A value below 1 indicates a lower estimated event hazard in the brigatinib group relative to crizotinib under the fitted model.
The 95% CI of 0.35–0.66 indicates that the point estimate should be interpreted together with a range of plausible values under the specified statistical framework. The confidence interval is not a prediction interval for individual participants.
The reported P-value < 0.0001 indicates strong evidence against the null hypothesis in the reported primary statistical comparison. The P-value should not be used to determine the size of the treatment effect.
The ClinicalTrials.gov record does not report median PFS, absolute PFS probabilities at particular time points, or individual participant event histories. Those quantities cannot be reconstructed from the HR and confidence interval alone.
28. Related Tutorials
Learn more about the methods used in this trial:
29. Related Calculators
30. Sources
- ClinicalTrials.gov: NCT02737501 — ALTA-1L.
- PubMed: PMID 41740899.
- PubMed: PMID 41410381.
- PubMed: PMID 39707229.
- PubMed: PMID 32780660.
- PubMed: PMID 30280657.
Continue through Clinical Biostats
Connect this trial's endpoints and statistical methods with deeper tutorials, statistical calculators, and other clinical-trial analyses.
31. Record Summary
ALTA-1L provides a compact example of how different clinical-trial endpoints require different statistical tools. The primary PFS endpoint is a time-to-event measure analyzed with a log-rank test and a Cox proportional-hazards model. The posted response and disease-control analyses use stratified Cochran-Mantel-Haenszel methods and odds ratios, while global health status and quality of life use a mixed-effects model with an unstructured covariance matrix.
The primary PFS estimate was HR 0.481 with a two-sided 95% CI of 0.35–0.66 and P < 0.0001. Secondary analyses reported an OR of 1.74 for confirmed ORR, an OR of 13.56 for confirmed intracranial ORR, an HR of 0.293 for intracranial PFS, an OR of 0.93 for DCR, and a least square mean difference of 5.79 for the specified global health status/quality-of-life outcome.
The most important statistical lesson is that these estimates should remain tied to their endpoint definitions and analysis methods. Hazard ratios, odds ratios, and least square mean differences answer different questions. Confidence intervals communicate precision, P-values communicate evidence against specified null hypotheses, and the ITT population preserves the randomized comparison. Reading all of these components together provides a more rigorous interpretation than relying on any single numerical result.