This page separates reported trial results from statistical interpretation. Trial-specific numerical results are restricted to the ClinicalTrials.gov record. Where the registry does not provide an estimate, confidence interval, or named statistical method, this page does not infer one from another source.
1. Trial at a Glance
DASISION was a randomized, open-label, parallel phase 3 study comparing dasatinib with imatinib in patients with newly diagnosed chronic-phase chronic myeloid leukemia. The registry reports 547 participants, two treatment arms, one registered primary endpoint, nine posted outcome measures, and four posted statistical analyses.
| Feature | DASISION |
|---|---|
| Trial name | DASISION |
| Phase | Phase 3 |
| Condition | Myeloid Leukemia, Chronic |
| Population | Patients with newly diagnosed chronic-phase chronic myeloid leukemia |
| Design | Randomized, parallel-group |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 547 |
| Interventions | Dasatinib and imatinib |
| Primary endpoint | Number of Participants With Best Confirmed Complete Cytogenetic Response (cCCyR) Within 12 Months |
| Primary endpoint type | Binary |
| Hypothesis type | Superiority |
| Primary analysis method | Cochran-Mantel-Haenszel test, stratified by Hasford Score |
| Trial status | Completed |
| Start | September 2007 |
| Primary completion | December 2009 |
2. Clinical Question
The central statistical question was whether the proportion of randomized participants achieving a best confirmed complete cytogenetic response (cCCyR) within 12 months differed between dasatinib and imatinib, with the comparison evaluated under a superiority hypothesis.
Population
Patients with newly diagnosed chronic-phase chronic myeloid leukemia.
Intervention
Dasatinib.
Comparator
Imatinib.
Primary question
Does the distribution of the binary primary response endpoint differ between randomized treatment groups after stratification by Hasford Score?
3. Trial Design
Dasatinib
- Drug intervention.
- Compared with imatinib under randomized allocation.
- Included in the primary cCCyR comparison.
Imatinib
- Drug intervention.
- Comparator for the randomized treatment comparison.
- Included in the primary cCCyR comparison.
The ClinicalTrials.gov record does not provide arm-specific enrollment counts. The safety analysis does, however, identify 258 participants at risk in each arm for the reported serious-adverse-event comparison. That safety denominator should not be substituted for the overall randomized enrollment unless the relevant analysis population is explicitly defined that way.
4. Primary Endpoint
| Endpoint | Registry definition / time frame | Analysis |
|---|---|---|
| Number of Participants With Best Confirmed Complete Cytogenetic Response (cCCyR) Within 12 Months | Pretreatment, every 3 months up to 12 months | Cochran-Mantel-Haenszel test, stratified by Hasford Score |
What counts as cCCyR?
The registry defines cytogenetic response as being based on the prevalence of Philadelphia positive (Ph+) cells in metaphase from a bone marrow sample. Ideally 25 metaphases, but at least 20 metaphases, were evaluated. Complete cytogenetic response (CCyR) is defined as 0% Ph+ cells in metaphase in bone marrow.
5. Primary Results
Best Confirmed Complete Cytogenetic Response Within 12 Months
Cochran-Mantel-Haenszel test
Analysis population: all randomized participants
Comparison: dasatinib vs imatinib · Hypothesis: superiority
Stratification factor: Hasford Score
| Primary endpoint | Analysis population | Method | P-value |
|---|---|---|---|
| Number of Participants With Best Confirmed Complete Cytogenetic Response (cCCyR) Within 12 Months | All randomized participants | Cochran-Mantel-Haenszel test stratified by Hasford Score | 0.0056 |
The reported P = 0.0056 is the result of the prespecified registry-posted primary comparison using the Cochran-Mantel-Haenszel test stratified by Hasford Score. Because the hypothesis type is superiority, the analysis was designed to test whether the treatment groups differed with respect to the binary cCCyR endpoint.
The p-value does not quantify the magnitude of the treatment effect. It does not tell us how much larger the cCCyR rate was in either arm, and it is not a probability that the null hypothesis is true.
A confidence interval for a risk difference, risk ratio, odds ratio, or another binary-effect measure would provide information about the magnitude and precision of the treatment difference. Those quantities are different from the p-value and would answer a different part of the statistical question.
The use of the Cochran-Mantel-Haenszel procedure is important because the primary comparison was stratified by Hasford Score rather than treated as an entirely unadjusted comparison. The interpretation therefore concerns the treatment comparison after accounting for the registered stratification factor in the analysis.
Why a Binary Endpoint Changes the Analysis
The primary endpoint is binary: a participant either meets the registry definition of best confirmed cCCyR within the specified assessment window or does not. That makes this analysis structurally different from the secondary endpoints expressed as time-to-event measures.
Possible effect measures include a risk difference, risk ratio, or odds ratio. The posted DASISION analysis instead identifies the Cochran-Mantel-Haenszel test and reports its p-value.
6. Secondary Endpoint Results
The registry contains three posted secondary statistical analyses. These use hazard ratios for time-to-event outcomes. The ClinicalTrials.gov record provides point estimates and two-sided 95% confidence intervals for each of these analyses.
Percentage of Participants Remaining in Confirmed Complete Cytogenetic Response
Hazard ratio
95% CI: 0.55–1.13
Time frame: Years 2, 3, 4 and 5
This analysis included all randomized participants who achieved cCCyR. The registry identifies the endpoint as time-to-event and the effect measure as a hazard ratio, but the ClinicalTrials.gov record does not report the specific statistical method used to estimate the hazard ratio.
A hazard ratio of 0.79 is a relative time-to-event measure. In a conventional event-time interpretation, a value below 1 indicates a lower estimated instantaneous event rate in the numerator treatment group relative to the comparator, subject to the event definition and model used. The registry's the ClinicalTrials.gov record does not state enough about the underlying event definition or model to attach a more specific model-based interpretation than that.
The 95% confidence interval of 0.55–1.13 expresses uncertainty around the estimated hazard ratio. Because the interval includes 1, the registry-reported interval does not exclude a hazard ratio of 1 at the 95% confidence level. That is an uncertainty statement, not a statement that the treatments are identical.
The hazard ratio does not describe the proportion of participants who remained in response, nor does it directly describe an absolute difference in percentages at a particular year. Those are different estimands.
The registry does not provide a formal p-value for this secondary analysis in the ClinicalTrials.gov record. A p-value should therefore not be reconstructed from the confidence interval.
Time to Confirmed Complete Cytogenetic Response Overall
Hazard ratio
95% CI: 1.20–1.77
Time frame: Day 1 to 5 years
The analysis concerns all randomized participants who achieved cCCyR. The registry notes that the hazard ratio and confidence interval were based on analyses on all randomized subjects.
A hazard ratio of 1.46 indicates an estimated event rate 1.46 times that of the comparator under the time-to-event analysis represented by the registry result. Equivalently, the point estimate is above 1, indicating a higher estimated instantaneous rate of the event for the numerator group in the reported comparison.
The 95% confidence interval of 1.20–1.77 lies above 1. The interval therefore indicates that the data are not consistent with a hazard ratio of exactly 1 under the stated 95% confidence framework.
Importantly, the hazard ratio is not a ratio of median times, and it should not be converted into a statement such as "the response occurred 46% sooner" without additional assumptions. A hazard ratio compares instantaneous event rates, not simply elapsed times.
The registry does not report the underlying statistical method for this analysis in the ClinicalTrials.gov record. Accordingly, this page does not assign a Cox model, log-rank test, Kaplan-Meier estimator, or another specific method to the posted result.
Time to Major Molecular Response (MMR) Overall
Hazard ratio
95% CI: 1.25–1.89
Time frame: Day 1 to 5 years
The analysis population is described in the registry as all participants who received treatment and achieved MMR. The analysis record again notes that the hazard ratio and confidence interval were based on analyses on all randomized subjects.
The point estimate of 1.54 is above 1, indicating a higher estimated instantaneous rate of the reported event for the numerator treatment group relative to the comparator. The 95% confidence interval, 1.25–1.89, also remains above 1.
As with the other hazard ratio, this does not mean that the time to MMR was exactly 54% shorter or that every participant experienced the same relative change. A hazard ratio is a relative event-rate measure across follow-up.
The confidence interval provides information about precision. It is narrower than a very imprecise interval might be, but it still represents uncertainty about the underlying treatment effect. The registry does not provide a formal p-value for this analysis in the ClinicalTrials.gov record.
7. Putting the Primary and Secondary Analyses Together
The DASISION registry record illustrates an important distinction in clinical-trial statistics: the primary endpoint and secondary time-to-event endpoints are not analyzed on the same statistical scale.
| Endpoint | Type | Population | Effect / test reported | Result |
|---|---|---|---|---|
| Best confirmed cCCyR within 12 months | Binary | All randomized participants | Cochran-Mantel-Haenszel test, stratified by Hasford Score | P = 0.0056 |
| Participants remaining in confirmed cCCyR | Time-to-event | All randomized participants who achieved cCCyR | Hazard ratio | 0.79 (95% CI 0.55–1.13) |
| Time to confirmed cCCyR overall | Time-to-event | All randomized participants who achieved cCCyR | Hazard ratio | 1.46 (95% CI 1.20–1.77) |
| Time to MMR overall | Time-to-event | All participants who received treatment and achieved MMR | Hazard ratio | 1.54 (95% CI 1.25–1.89) |
The direction of a hazard ratio must always be interpreted in relation to the exact event being analyzed. A hazard ratio below 1 is not intrinsically "better," and a hazard ratio above 1 is not intrinsically "worse." The meaning depends on whether the event represents a favorable event, an unfavorable event, progression, response, loss of response, or another endpoint.
This is particularly important for DASISION because the secondary endpoints include both time to a favorable response and remaining in a favorable response. The same numerical direction can have different substantive meanings depending on which event is being modeled.
8. Statistical Methodology
Cochran-Mantel-Haenszel Test
The primary analysis used the Cochran-Mantel-Haenszel (CMH) test, stratified by Hasford Score. The CMH framework is designed for comparing categorical outcomes between treatment groups while accounting for a categorical stratification variable.
Rather than treating all observations as belonging to a single unstratified 2 × 2 table, the CMH framework evaluates the treatment-response relationship across strata and combines the information into an overall test.
Why stratification matters
Randomization establishes the treatment comparison, but the analysis can still incorporate prespecified stratification factors. Here, the registry explicitly identifies Hasford Score as the stratification variable used in the primary CMH test.
The practical implication is that the primary test is not simply an unadjusted comparison of two overall response proportions. The analysis recognizes the stratified structure represented by Hasford Score.
Binary endpoint analysis
The primary endpoint is a binary response measured within a defined period. Binary endpoints are often summarized by the proportion of participants meeting the response definition in each group, followed by an effect measure and hypothesis test.
The registry analysis reports the hypothesis test but does not provide the corresponding response percentages, risk difference, risk ratio, odds ratio, or confidence interval. A statistically complete effect-size presentation would ordinarily benefit from those quantities, but they cannot be added here without violating the trial-data restriction.
Time-to-event analysis
The three secondary analyses are explicitly classified as time-to-event endpoints and report hazard ratios. Time-to-event methods account for the timing of events and can accommodate participants whose event status is not observed over the full follow-up period through censoring mechanisms.
The formula is conceptual. The exact model used to produce the posted DASISION secondary hazard ratios is not reported in the ClinicalTrials.gov record.
Confidence intervals
A 95% confidence interval describes the statistical uncertainty associated with an estimated effect under the assumptions of the relevant inferential procedure. It should be read together with the point estimate rather than as a substitute for it.
For example, an HR of 0.79 with a 95% CI of 0.55–1.13 communicates both the estimated relative effect and the range of values compatible with the specified confidence framework. The interval includes 1, so the analysis does not exclude no relative hazard difference at that confidence level.
9. Statistical Methods Explained
Why was a Cochran-Mantel-Haenszel test used?
The registry identifies the primary endpoint as binary and the primary method as the Cochran-Mantel-Haenszel test. This method is appropriate for categorical treatment comparisons when the analysis incorporates a stratification variable. In DASISION, the registered analysis was stratified by Hasford Score.
What does stratification by Hasford Score mean?
Stratification means that the treatment-response comparison is evaluated while preserving the structure defined by the Hasford Score strata. The resulting CMH test combines information across those strata rather than ignoring the registered stratification factor.
What does P = 0.0056 mean?
It is the reported p-value from the primary CMH test. Under the null hypothesis represented by that test and its assumptions, it quantifies how incompatible the observed stratified categorical data are with the null comparison. It does not measure the size of the treatment effect and is not the probability that the null hypothesis is true.
What does a hazard ratio of 1.46 mean?
A hazard ratio of 1.46 means that the estimated instantaneous event rate for the numerator group is 1.46 times that of the comparator under the relevant time-to-event analysis. It does not mean that the median event time differs by 46%, and it does not mean that 46% more participants experienced the event.
Why does the confidence interval matter?
The point estimate alone gives only one estimate of the relative effect. The confidence interval indicates how precisely that estimate was measured under the statistical framework. For the reported HR of 1.46, the 95% CI is 1.20–1.77; for the reported HR of 0.79, it is 0.55–1.13. These intervals lead to different conclusions about whether a hazard ratio of 1 is excluded.
Why can two hazard ratios not be interpreted without knowing the event?
A hazard ratio is directional only relative to its event definition. A larger hazard of achieving a favorable response can be desirable, whereas a larger hazard of losing that response could be undesirable. Therefore, the event itself must be identified before the numerical direction of a hazard ratio can be given substantive meaning.
10. Analysis Populations and Estimands
The primary analysis population is explicitly stated as all randomized participants. This is important because randomization is the mechanism that creates the intended comparability between treatment groups at the start of the trial.
| Analysis | Population specified in the registry |
|---|---|
| Primary cCCyR analysis | All randomized participants |
| Remaining in confirmed cCCyR | All randomized participants who achieved cCCyR |
| Time to confirmed cCCyR | All randomized participants who achieved cCCyR |
| Time to MMR | All participants who received treatment and achieved MMR |
These populations are not interchangeable. The primary analysis begins with the randomized population, whereas several secondary analyses condition on achieving a response. Conditioning on a post-randomization outcome can change the population being compared and therefore changes the scientific question.
For example, an analysis among participants who achieved cCCyR asks a different question from the primary analysis of achieving cCCyR among all randomized participants. The former describes timing or durability among responders; the latter evaluates the randomized treatment comparison for attaining the response.
11. Time Frames and Endpoint Structure
Pretreatment through 12 months
The primary endpoint was assessed pretreatment and every 3 months up to 12 months.
Years 2, 3, 4 and 5
The percentage of participants remaining in confirmed cCCyR was evaluated at Years 2, 3, 4 and 5.
Day 1 to 5 years
Time to confirmed cCCyR overall was evaluated from Day 1 to 5 years.
Day 1 to 5 years
Time to MMR overall was evaluated from Day 1 to 5 years.
The different time frames are not merely administrative details. A binary endpoint measured within 12 months and a time-to-event endpoint followed through 5 years estimate different features of treatment response. They should not be combined into a single numerical interpretation.
12. Serious Adverse Events
The ClinicalTrials.gov record reports serious adverse events by arm using affected participants over participants at risk.
| Safety measure | Dasatinib | Imatinib |
|---|---|---|
| Serious adverse events | 90 / 258 | 70 / 258 |
The ClinicalTrials.gov record does not include a formal hypothesis test, confidence interval, exposure adjustment, or statistical model for serious adverse events. Accordingly, this page reports the affected/at-risk counts without constructing an unreported inferential comparison.
The reported counts show that serious adverse events affected 90 of 258 participants in the dasatinib group and 70 of 258 in the imatinib group in the registry-reported safety analysis. The difference in counts should not automatically be interpreted as a statistically significant treatment difference because no inferential analysis for this safety endpoint is provided in the ClinicalTrials.gov record.
Safety and efficacy also answer different questions. A response endpoint asks whether participants achieved a defined clinical or biological outcome, whereas a serious-adverse-event analysis asks about harm. They should be presented separately rather than collapsed into a single numerical benefit-risk score.
13. Multiplicity, Interim Analysis, and Other Design Features
The ClinicalTrials.gov record identifies the primary hypothesis as superiority and identify one registered primary endpoint. They do not provide a non-inferiority margin, factorial design, crossover specification, interim-analysis plan, alpha-spending method, multiplicity-adjustment procedure, or Bayesian method.
| Design topic | What the registry-reported DASISION data support |
|---|---|
| Superiority | Yes. The primary and secondary analyses are identified with a superiority hypothesis. |
| Non-inferiority margin | Not reported in the ClinicalTrials.gov record. |
| Crossover | Not reported in the ClinicalTrials.gov record. |
| Factorial design | Not reported; the registered design model is parallel. |
| Multiplicity procedure | Not reported in the ClinicalTrials.gov record. |
| Interim analysis | Not reported in the ClinicalTrials.gov record. |
| Missing-data/imputation method | Not reported in the ClinicalTrials.gov record. |
| Bayesian methods | Not reported in the ClinicalTrials.gov record. |
| Primary stratification | Hasford Score for the CMH primary analysis. |
This distinction is important. Absence of a method in the ClinicalTrials.gov record is not evidence that the investigators did not use such a method elsewhere in the trial's full statistical documentation. It means only that this page does not attribute an unreported procedure to DASISION.
14. Why the Primary Endpoint and Hazard Ratios Should Be Read Separately
The primary endpoint is a binary outcome observed during a specified 12-month assessment window. The secondary analyses use time-to-event estimands. Although both can compare the same randomized treatment groups, they encode different information.
Binary response
Answers whether a participant achieved the defined best confirmed cCCyR within the primary time window.
Time to response
Uses the timing of the confirmed cCCyR event rather than only whether it occurred within a fixed window.
Remaining in response
Describes a time-to-event process among participants who achieved cCCyR.
Time to MMR
Uses another molecular-response endpoint and therefore represents a distinct biological outcome from cCCyR.
This separation prevents a common statistical mistake: assuming that all reported treatment effects are interchangeable. A p-value for a binary endpoint, a hazard ratio for time to response, and a hazard ratio for time to molecular response are different statistical quantities.
15. Hazard Ratios: Interpretation and Cautions
A hazard ratio compares estimated instantaneous event rates. It is therefore not automatically equivalent to a relative risk, odds ratio, probability ratio, or percentage-point difference.
For DASISION, the registry-reported secondary endpoints involve confirmed cCCyR and MMR. The direction of the HR must therefore be interpreted with the exact event definition and analysis population in view.
The HR point estimate is only one estimate of the treatment contrast. The accompanying 95% CI indicates uncertainty around that estimate. For example, 0.79 with a 95% CI of 0.55–1.13 communicates substantially different information from a point estimate alone.
When a Cox proportional-hazards model is used, a constant hazard ratio over time is a key modeling assumption. The registry-reported DASISION registry data do not identify the statistical model used for the posted secondary hazard ratios, so this page does not assert that a Cox model was used. More generally, however, any interpretation of a single HR should consider whether a constant relative hazard is a reasonable summary of the event process.
16. Limitations of the Posted Statistical Record
- Incomplete effect-size reporting for the primary endpoint: the registry-reported primary analysis reports the CMH method and P-value 0.0056 but does not provide the treatment-specific cCCyR response counts or percentages, nor a confidence interval for an effect measure.
- Incomplete method reporting for secondary hazard ratios: the registry analyses provide hazard ratios and confidence intervals but do not identify the specific statistical method used to generate them.
- No secondary p-values reported: the three secondary analyses provide point estimates and 95% confidence intervals but no formal p-values in the ClinicalTrials.gov record.
- Different analysis populations: the primary analysis uses all randomized participants, whereas the secondary response-related analyses condition on response or treatment receipt as specified in the registry.
- No baseline table in the ClinicalTrials.gov record: detailed baseline demographic or disease characteristics are not available in the ClinicalTrials.gov record and are therefore not reproduced here.
- No subgroup estimates in the ClinicalTrials.gov record: the ClinicalTrials.gov record does not include subgroup-specific treatment effects.
- No missing-data procedure is specified: the ClinicalTrials.gov record does not describe imputation, censoring rules, or other missing-data handling procedures.
- No interim-analysis information is reported: the statistical record provided here does not identify an interim monitoring strategy or alpha-spending approach.
- Safety inference is limited: serious-adverse-event counts are available, but no formal inferential analysis is posted on ClinicalTrials.gov for that safety endpoint.
17. Why This Trial Matters Statistically
DASISION is a useful statistical teaching case because the registry record connects a randomized superiority comparison with both a stratified binary endpoint analysis and several time-to-event analyses. It therefore illustrates how one randomized trial can require multiple statistical frameworks depending on the endpoint.
| Concept | How it appears in DASISION |
|---|---|
| Randomization | The trial uses randomized allocation in a parallel-group design. |
| Superiority testing | The primary endpoint is analyzed under a superiority hypothesis. |
| Binary endpoints | The primary cCCyR endpoint is classified as binary. |
| Cochran-Mantel-Haenszel test | The primary comparison uses a CMH test stratified by Hasford Score. |
| Stratified analysis | Hasford Score is incorporated into the primary statistical comparison. |
| Confidence intervals | Secondary hazard-ratio analyses report two-sided 95% confidence intervals. |
| Hazard ratios | Three posted secondary analyses report hazard ratios. |
| Time-to-event endpoints | Remaining in cCCyR, time to cCCyR, and time to MMR are classified as time-to-event outcomes. |
| Analysis populations | The primary endpoint uses all randomized participants, while secondary analyses use specified response/treatment populations. |
| Safety analysis | Serious adverse events are reported as affected participants over participants at risk by arm. |
The most important educational lesson is that statistical interpretation must follow the endpoint. The primary cCCyR analysis is a stratified categorical comparison, whereas the secondary HR analyses are time-to-event results. Treating every result as though it were the same type of effect would obscure rather than clarify the statistical evidence.
18. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
19. Related Statistical Calculators
20. Sources
- ClinicalTrials.gov: DASISION, NCT00481247.
- PubMed record: PMID 30120281.
- PubMed record: PMID 26118315.
- PubMed record: PMID 24357015.
- PubMed record: PMID 24311723.
- PubMed record: PMID 22160483.
Continue through the Clinical Biostats statistical tutorials
Explore the statistical concepts behind randomized trials, categorical endpoints, confidence intervals, hazard ratios, stratified analyses, and time-to-event methods.
21. Record Summary
DASISION is a randomized phase 3 trial of dasatinib versus imatinib in patients with newly diagnosed chronic-phase chronic myeloid leukemia. The ClinicalTrials.gov record identifies a binary primary endpoint—best confirmed complete cytogenetic response within 12 months—and analyzes it using a Cochran-Mantel-Haenszel test stratified by Hasford Score. The reported primary P-value is 0.0056.
The registry also reports three secondary time-to-event analyses. Remaining in confirmed cCCyR has a reported HR of 0.79 with a 95% CI of 0.55–1.13; time to confirmed cCCyR overall has an HR of 1.46 with a 95% CI of 1.20–1.77; and time to MMR overall has an HR of 1.54 with a 95% CI of 1.25–1.89.
The central statistical lesson is that these results represent different estimands. The primary CMH test addresses a stratified binary treatment comparison, while the secondary hazard ratios describe time-to-event relationships. Correct interpretation therefore requires attention to the endpoint definition, analysis population, time frame, effect measure, confidence interval, and statistical method actually reported.