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
VIALE-A is a phase 3 randomized, parallel-group trial in acute myeloid leukemia. The registry describes a comparison involving azacitidine with placebo versus venetoclax plus azacitidine, with primary analyses of overall survival and complete remission/complete remission with incomplete marrow recovery.
| Feature | VIALE-A |
|---|---|
| Trial name | VIALE-A |
| ClinicalTrials.gov identifier | NCT02993523 |
| Phase | Phase 3 |
| Condition | Acute Myeloid Leukemia (AML) |
| Population | Treatment-naive participants with acute myeloid leukemia who are ineligible for standard induction therapy |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 443.0 |
| Trial status | Active, not recruiting |
| Start | 2 February 2017 |
| Primary completion | 1 December 2021 |
| Lead sponsor | AbbVie |
| Sponsor type | Industry |
2. Clinical Question
The primary statistical question was whether the venetoclax-plus-azacitidine regimen differed from placebo plus azacitidine with respect to overall survival and the percentage of participants with complete remission (CR) and complete remission with incomplete marrow recovery (CRi).
Population
Treatment-naive participants with acute myeloid leukemia who are ineligible for standard induction therapy.
Intervention
Venetoclax 100 mg/200 mg/400 mg in combination with azacitidine 75 mg/m2.
Comparator
Placebo plus azacitidine 75 mg/m2.
Primary question
Does the venetoclax-plus-azacitidine regimen demonstrate superiority on the registered primary endpoints?
3. Trial Design
Placebo + azacitidine
- Placebo
- Azacitidine 75 mg/m2
- Included in the full analysis set for the primary efficacy comparison
Venetoclax + azacitidine
- Venetoclax 100 mg/200 mg/400 mg
- Azacitidine 75 mg/m2
- Included in the full analysis set for the primary efficacy comparison
4. Endpoints
| Primary endpoint | Registry definition / time frame | Endpoint type |
|---|---|---|
| Overall Survival (OS) | OS is defined as the number of days from the date of randomization to the date of death. Time frame: from the study start up to death or alive or lost to follow-up, up to approximately 4.8 years; data cut-off date: 1 December 2021. | Time-to-event |
| Percentage of Participants With Complete Remission (CR) and Complete Remission With Incomplete Marrow Recovery (CRi) | CR and CRi was calculated based on current International Working Group (IWG) criteria. CR is defined as absolute neutrophil count >103/microliter (mcL), platelets >105/mcL, red cell transfusion independence, and bone marrow with <5% blasts. CRi is defined as bone marrow with less than 5% blasts, and absolute neutrophils of ≤103/mcL or platelets ≤105/mcL. Time frame: from the study start up to death, up to approximately 4.8 years; data cut-off date: 1 December 2021. | Binary |
The registry identifies two primary endpoint types: one time-to-event endpoint and one binary endpoint. That distinction drives the choice of statistical test and effect measure.
5. Statistical Methodology
Overall survival: log-rank test and Cox proportional-hazards model
The registry states that a log-rank test was used to compare the OS distributions between the two treatment arms. The hazard ratio was reported from a Cox regression model. The posted analysis specifies that the Cox model was stratified by age and cytogenetic risk.
The posted analysis specifies stratification by age (18 - < 75, ≥ 75) and cytogenetic risk (intermediate, poor).
CR/CRi: Cochran-Mantel-Haenszel test
The binary primary endpoint was analyzed using the Cochran-Mantel-Haenszel (CMH) test. The registry specifies that the CMH test was stratified by age (18 - < 75, ≥ 75) and cytogenetic risk (intermediate, poor).
The CMH framework is appropriate when a binary outcome is compared across two treatment groups while accounting for prespecified strata. Instead of treating all observations as if they came from a single homogeneous population, the stratified procedure combines information across the defined strata.
Intention-to-treat analysis
The posted primary analyses identify intention-to-treat analysis as a concept in the analysis text. For OS, the full analysis set included all Group 2 participants randomized by IVRS/IWRS and excluded the open-label China safety cohort. This anchors the primary efficacy comparison to randomized treatment assignment rather than subsequent treatment exposure.
Stratification
Both primary analyses were stratified by the same two factors: age and cytogenetic risk. Stratification can improve the alignment between the statistical analysis and the randomized trial structure when those factors are prespecified as important sources of variation.
| Endpoint | Analysis | Effect measure | Stratification |
|---|---|---|---|
| Overall survival | Log-rank test; Cox proportional-hazards model | Hazard ratio | Age and cytogenetic risk |
| CR + CRi | Cochran-Mantel-Haenszel test | No estimate reported in the posted statistical analysis | Age and cytogenetic risk |
6. Results: Overall Survival
The registry reports a formal primary analysis of overall survival in the full analysis set. The comparison was between placebo plus azacitidine 75 mg/m2 and venetoclax 100 mg/200 mg/400 mg plus azacitidine 75 mg/m2.
Hazard ratio for death
95% CI: 0.465–0.723 · P < 0.001
Two-sided 95% confidence interval; superiority hypothesis.
| Endpoint | Comparison | Method | Effect estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Overall Survival | Venetoclax + azacitidine vs placebo + azacitidine | Log-rank test; stratified Cox model | HR 0.580 | 0.465–0.723 | <0.001 |
An HR of 0.580 means that, under the fitted Cox model, the estimated instantaneous hazard of death in the venetoclax-plus-azacitidine group was 0.580 times the corresponding hazard in the placebo-plus-azacitidine group. Expressed as a simple relative interpretation, this corresponds to an estimated 42% lower hazard because 1 − 0.580 = 0.420.
The HR does not mean that 42% of participants avoided death, that survival increased by 42%, or that each individual participant experienced exactly a 42% reduction in risk. A hazard ratio is a relative time-to-event measure derived from a statistical model.
The two-sided 95% CI of 0.465–0.723 describes uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of outcomes that individual patients might experience.
The P < 0.001 result addresses the evidence against the null hypothesis used for the statistical comparison. It does not measure the size of the treatment effect. The HR and its confidence interval provide the effect-size and precision information.
Because the estimate comes from a Cox proportional-hazards model, interpretation of a single HR also depends on the model's proportional-hazards framework. The registry does not provide additional information here about whether that assumption was formally assessed.
What the OS analysis establishes statistically
The posted analysis provides three complementary pieces of information: a prespecified time-to-event comparison using the log-rank test, a hazard-ratio estimate from a stratified Cox model, and a two-sided 95% confidence interval around that estimate. Together, these describe both the direction and precision of the estimated relative treatment effect.
The analysis population is also important. The registry specifies that the full analysis set contained all Group 2 participants randomized by IVRS/IWRS and excluded the open-label China safety cohort. Therefore, the posted OS estimate should not be generalized to the China safety cohort as though it were part of the same randomized efficacy comparison.
7. Results: Complete Remission and Complete Remission With Incomplete Marrow Recovery
The second primary endpoint was the percentage of participants with CR and CRi. The registry reports a formal CMH analysis and a superiority hypothesis.
Stratified comparison
Cochran-Mantel-Haenszel test
Stratified by age (18 - < 75, ≥ 75) and cytogenetic risk (intermediate, poor).
| Endpoint | Method | Analysis population | Estimate | P-value |
|---|---|---|---|---|
| Percentage of participants with CR and CRi | Cochran-Mantel-Haenszel test | Full Analysis Set; open-label China safety cohort excluded from the primary randomized comparison | Not reported in the posted statistical analysis | <0.001 |
The reported P < 0.001 indicates a statistically significant difference under the prespecified CMH comparison and superiority hypothesis. It does not by itself quantify the magnitude of that difference.
Unlike the OS analysis, the registry's posted statistical-analysis record does not provide an effect estimate or confidence interval for this binary endpoint. Therefore, a numerical treatment effect, absolute percentage-point difference, odds ratio, or relative risk should not be reconstructed from the ClinicalTrials.gov record.
The stratified CMH test accounts for the specified age and cytogenetic-risk strata. Its P-value should therefore be interpreted as the result of that stratified comparison rather than as an unadjusted two-group test.
The endpoint itself is also more specific than a generic "response" outcome: the registry defines CR and CRi using International Working Group criteria. The statistical endpoint is the percentage of participants meeting the combined CR/CRi definition.
8. Serious Adverse Events
The registry data provide serious adverse event counts as affected participants over participants at risk. These figures are presented separately from the primary efficacy analyses because safety and efficacy answer different questions and use different analysis populations and objectives.
| Safety group | Affected / at risk | Registry description |
|---|---|---|
| Placebo + azacitidine | 111 / 144 | Group 1 and Group 2 |
| Venetoclax 100 mg/2 | 242 / 283 | Group 1 and Group 2 |
| Venetoclax 400 | 7 / 10 | Open Label China Cohort |
The registry data therefore identify serious adverse event experience in the randomized-treatment groups and separately in the open-label China cohort. Because the ClinicalTrials.gov record does not provide a formal comparative statistical test for these serious adverse event counts, they should not be converted into a treatment-effect claim beyond the reported affected and at-risk counts.
9. Statistical Methods Explained
Why was a log-rank test used for overall survival?
Overall survival is a time-to-event endpoint because participants can experience the event at different times and some participants may be alive or lost to follow-up at the time of analysis. The log-rank test compares the observed event pattern between treatment groups across follow-up while accommodating right censoring.
What does an OS hazard ratio of 0.580 mean?
The HR of 0.580 is a relative measure from the Cox proportional-hazards model. Within that model, the estimated instantaneous death hazard in the venetoclax-plus-azacitidine group is 58.0% of that in the comparator group. It is not a probability of survival and is not an absolute risk difference.
Why is the confidence interval important?
The 95% CI of 0.465–0.723 provides a range describing statistical uncertainty around the estimated HR. A point estimate alone can conceal how precisely the treatment effect has been estimated. The confidence interval also helps distinguish the magnitude of the observed estimate from the uncertainty surrounding it.
Why does the P-value not measure effect size?
A P-value describes the compatibility of the observed data with the specified null hypothesis under the statistical testing framework. It is affected by both the magnitude of an effect and the amount of information available. The HR describes relative effect size; the confidence interval describes precision. These quantities answer different questions.
Why was the Cochran-Mantel-Haenszel test used for CR/CRi?
CR/CRi is a binary endpoint: participants either meet the combined response definition or they do not. A CMH test allows the treatment comparison to be performed while accounting for the prespecified age and cytogenetic-risk strata rather than ignoring those strata.
Why does stratification matter?
Stratification separates the comparison into prespecified categories and then combines the evidence across them according to the stratified statistical method. In VIALE-A, both primary analyses identify age and cytogenetic risk as the stratification factors.
Why is the analysis population important?
The OS analysis specifies a full analysis set containing all Group 2 participants randomized by IVRS/IWRS while excluding the open-label China safety cohort. Keeping the analysis population aligned with the randomized comparison helps preserve the interpretability of the treatment contrast.
10. Understanding the Time-to-Event Analysis
Kaplan-Meier estimation
Although the registry's posted OS analysis identifies the log-rank test and Cox regression as the formal methods, Kaplan-Meier estimation is the standard descriptive framework for displaying time-to-event distributions. It accounts for participants who contribute follow-up before being censored.
Here, di represents the number of events at an event time and ni represents the number at risk immediately before that time.
The important point is that a survival analysis uses both event timing and censoring information. A binary endpoint measured once does not contain the same temporal information.
Log-rank testing
The log-rank test compares the survival experience of two groups over the follow-up period. In VIALE-A, the registry explicitly identifies this as the method used to compare the OS distributions between the two treatment arms.
Cox proportional-hazards modeling
The Cox model provides a relative hazard estimate while allowing the baseline hazard to remain unspecified. VIALE-A's posted OS analysis reports the HR from a Cox model stratified by age and cytogenetic risk.
An HR is not equivalent to a risk ratio, odds ratio, probability, or absolute difference. Its interpretation is tied to the time-to-event model.
11. Stratified Analysis in VIALE-A
The same two stratification factors appear in both posted primary analyses. This provides an especially useful example of how a clinical-trial statistical plan can use the same prespecified factors across different endpoint types.
| Stratification factor | Categories reported in the analysis | Primary analyses using it |
|---|---|---|
| Age | 18 - < 75; ≥ 75 | OS and CR/CRi |
| Cytogenetic risk | Intermediate; poor | OS and CR/CRi |
For OS, stratification is incorporated into the Cox analysis and the trial's log-rank comparison. For CR/CRi, the CMH test uses the same strata. This distinction is educationally useful: stratification is a design and analysis principle, not a single statistical test.
12. Intention-to-Treat and the Full Analysis Set
The registry analysis text identifies intention-to-treat analysis as a concept for both primary endpoints. For the OS analysis, the full analysis set is explicitly defined as all Group 2 participants randomized by IVRS/IWRS, excluding the open-label China safety cohort.
Why analyze randomized assignment?
Analyzing participants according to randomized assignment preserves the comparison created by randomization and avoids redefining treatment groups according to events that occur after randomization.
Why exclude the China cohort?
The registry explicitly identifies the open-label China cohort separately from the Group 2 randomized efficacy population. It therefore does not belong in the same primary randomized comparison.
ITT reasoning is particularly important in a randomized trial because post-randomization events can be related to treatment, disease status, or other factors. Changing the analysis population based on such events can undermine the original treatment comparison.
13. What the Confidence Interval Adds to the OS Result
The OS HR of 0.580 is the central estimate from the reported Cox model.
The two-sided 95% CI of 0.465–0.723 shows the statistical uncertainty around that point estimate. It is narrower than an extremely imprecise estimate would be, but the interval still contains a range of plausible model-based effect sizes.
The confidence interval does not state that 95% of individual patients will experience an HR somewhere between 0.465 and 0.723. It concerns uncertainty in the estimated population-level treatment effect under the statistical model.
14. Multiplicity and the Two Primary Endpoints
The registry identifies two primary endpoints: OS and CR/CRi. Both have formal statistical analyses posted, and both are specified as superiority hypotheses.
| Primary endpoint | Endpoint type | Formal method | P-value reported | Effect estimate reported |
|---|---|---|---|---|
| Overall Survival | Time-to-event | Log-rank test; Cox model | <0.001 | HR 0.580; 95% CI 0.465–0.723 |
| CR + CRi | Binary | Cochran-Mantel-Haenszel test | <0.001 | Not reported in registry-reported statistical analysis |
Because two primary endpoints are present, interpretation of the overall confirmatory framework ordinarily depends on the prespecified multiplicity strategy. The ClinicalTrials.gov record does not provide an alpha-allocation or hierarchical testing rule, so no additional multiplicity procedure should be inferred.
15. Missing Data, Censoring, and Analysis Assumptions
Overall survival is inherently a time-to-event endpoint. Participants who are alive or lost to follow-up at the relevant analysis time may contribute follow-up information without having experienced the event. The registered OS definition explicitly includes participants who are alive or lost to follow-up.
The ClinicalTrials.gov record does not specify a particular missing-data imputation method for either primary endpoint. Therefore, no imputation procedure is attributed to the trial here.
Similarly, the ClinicalTrials.gov record does not report a formal assessment of the proportional-hazards assumption. The HR should therefore be interpreted as the result of the reported Cox model, with the usual model-based caution that a single HR is most naturally interpreted under a proportional-hazards framework.
16. Non-Inferiority, Crossover, and Bayesian Methods
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Non-inferiority margin | Not reported; the registered hypothesis type is superiority. |
| Crossover | Not reported in the ClinicalTrials.gov record. |
| Factorial design | Not reported; the design model is parallel. |
| Interim analysis | Not reported in the ClinicalTrials.gov record. |
| Missing-data imputation | Not reported in the ClinicalTrials.gov record. |
| Bayesian methods | Not reported in the statistical analyses posted on ClinicalTrials.gov. |
This distinction is important because the absence of a method in the ClinicalTrials.gov record is not evidence that a method was definitively absent from every trial document. It means that the method cannot be responsibly attributed to the VIALE-A analysis from the ClinicalTrials.gov record.
17. Trial Timeline
Study start
The registered study start date is 2 February 2017.
Randomized parallel design
The study is registered as a phase 3 randomized, parallel trial with quadruple masking and a primary purpose of treatment.
Primary completion and data cutoff
The registry lists primary completion on 1 December 2021, and the primary endpoint analyses use a data cut-off date of 1 December 2021.
Active, not recruiting
The ClinicalTrials.gov record lists the study status as active, not recruiting.
18. Limitations
- Limited numerical reporting for CR/CRi: the ClinicalTrials.gov record reports the CMH P-value but does not provide a treatment-effect estimate or confidence interval for the binary endpoint.
- Limited survival detail: the ClinicalTrials.gov record provides the OS HR and confidence interval but do not provide median OS, Kaplan-Meier event counts, survival probabilities at specified time points, or a digitized survival curve.
- Analysis-population specificity: the primary OS analysis excludes the open-label China safety cohort, so the randomized efficacy estimate should not be generalized to that cohort.
- Safety comparison: serious adverse event counts are reported by affected participants and participants at risk, but no formal comparative safety test is reported.
- Unreported design details: the ClinicalTrials.gov record does not specify an interim-analysis plan, multiplicity strategy, missing-data/imputation approach, or non-inferiority margin.
- Model assumptions: the OS HR is derived from a Cox proportional-hazards model, but the ClinicalTrials.gov record does not report a formal assessment of proportional hazards.
- Registry-level evidence: this page is deliberately constrained to the ClinicalTrials.gov record and listed PubMed records. It does not supplement missing numerical results from external publications.
19. Why This Trial Matters Statistically
VIALE-A provides a compact teaching example of how a modern randomized trial can use different statistical methods for different endpoint types while retaining a common stratification framework.
| Concept | How it appears in VIALE-A |
|---|---|
| Randomization | The trial is registered as randomized with 443 participants enrolled. |
| Parallel design | The design model is parallel. |
| Quadruple masking | The registry identifies the study as quadruple masked. |
| Time-to-event analysis | Overall survival is the time-to-event primary endpoint. |
| Log-rank test | Used to compare the OS distribution between the two primary treatment arms. |
| Hazard ratio | The OS treatment effect is reported as HR 0.580 with a two-sided 95% CI. |
| Cox regression | Used to report the OS hazard ratio, with stratification by age and cytogenetic risk. |
| Binary endpoint analysis | CR/CRi is analyzed as a binary outcome. |
| Cochran-Mantel-Haenszel test | Used for the stratified CR/CRi comparison. |
| Stratified analysis | Age and cytogenetic risk are used as stratification factors in both primary analyses. |
| Intention-to-treat principle | The posted analyses identify ITT as an analysis concept, with the OS full analysis set based on randomized Group 2 participants. |
| Superiority testing | Both primary analyses are identified as superiority hypotheses. |
| Endpoint-specific effect measures | The OS analysis provides an HR and CI, whereas the posted CR/CRi analysis provides a P-value without an estimate or CI. |
20. A Statistical Reading of the Two Primary Results
Overall survival
The OS analysis supplies a complete relative-effect result: HR 0.580, two-sided 95% CI 0.465–0.723, and P < 0.001. The HR is estimated from a stratified Cox model following a log-rank comparison.
CR/CRi
The CR/CRi analysis supplies a formal stratified CMH comparison with P < 0.001, but the posted analysis does not supply an effect estimate or confidence interval. The magnitude of the difference therefore cannot be quantified from the ClinicalTrials.gov record.
This difference illustrates an important reporting principle: statistical significance and effect-size reporting are not interchangeable. A P-value can support a hypothesis test, while an effect estimate and confidence interval are needed to understand the magnitude and precision of a treatment difference.
21. What the OS Hazard Ratio Does — and Does Not — Mean
An HR of 0.580 corresponds to an estimated 42% lower instantaneous hazard of death under the fitted model, calculated as 1 − 0.580.
The HR does not provide the percentage-point difference in the probability of death by a particular time. An absolute survival difference would require time-specific survival estimates, which are not reported in the ClinicalTrials.gov record.
The HR is a population-level relative measure from a statistical model. It should not be interpreted as meaning that every participant experiences the same proportional change in individual risk.
The P < 0.001 result indicates strong evidence against the null hypothesis under the reported testing framework. It does not tell the reader whether an effect is large, small, or clinically important. The HR and its confidence interval address that different question.
22. Why the Endpoint Definitions Matter
Clinical-trial statistics begin with precise endpoint definitions. The VIALE-A registry distinguishes a survival endpoint from a binary response endpoint, and the two analyses consequently use different statistical procedures.
| Feature | Overall survival | CR/CRi |
|---|---|---|
| Outcome structure | Time until death | Whether the participant meets the combined CR/CRi definition |
| Registry endpoint type | Time-to-event | Binary |
| Posted primary method | Log-rank test | Cochran-Mantel-Haenszel test |
| Model-based effect measure posted | Hazard ratio | None in registry-reported analysis |
| Confidence interval posted | 95% two-sided CI | None in registry-reported analysis |
| Stratification | Age and cytogenetic risk | Age and cytogenetic risk |
Using the same statistical method for both endpoints would discard information about the structure of the outcomes. Time-to-event analysis accounts for event timing and censoring; binary analysis focuses on whether a defined outcome occurred.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: NCT02993523 — VIALE-A.
- Linked PubMed record: PubMed 40913104.
- Linked PubMed record: PubMed 39133921.
- Linked PubMed record: PubMed 36007102.
- Linked PubMed record: PubMed 35829925.
- Linked PubMed record: PubMed 35696071.
Continue through the Clinical Biostats statistical library
Use the methods in this trial as a starting point for deeper tutorials on survival analysis, categorical-data methods, stratification, confidence intervals, and randomized clinical-trial analysis.
26. Record Summary
VIALE-A provides a useful statistical case study because its two primary endpoints require different analytical frameworks. Overall survival is analyzed as a time-to-event outcome using a log-rank comparison and a stratified Cox model, producing an HR of 0.580 with a two-sided 95% CI of 0.465–0.723 and P < 0.001. The CR/CRi endpoint is analyzed as a binary outcome using a stratified Cochran-Mantel-Haenszel test, with P < 0.001 reported but no effect estimate or confidence interval reported in the posted analysis.
The statistical interpretation should therefore keep several distinctions clear: a hazard ratio is not an absolute risk difference; a P-value is not an effect size; a confidence interval describes uncertainty rather than individual patient outcomes; and a stratified analysis accounts for prespecified factors rather than simply increasing the number of observations. The analysis population also matters, because the primary randomized efficacy comparison explicitly excludes the open-label China safety cohort.