This page separates reported trial results from statistical interpretation. All trial-specific numerical results presented here come from the ClinicalTrials.gov record. The registry provides the official trial record.
1. Trial at a Glance
ASPEN was a randomized, parallel, open-label phase 3 trial comparing BGB-3111 and ibrutinib in participants with Waldenström's macroglobulinemia. The primary endpoint was binary: the percentage of participants achieving either a complete response or very good partial response using an adaptation of response criteria updated at the Sixth International Workshop on WM, as assessed by an independent review committee.
| Feature | ASPEN |
|---|---|
| Trial name | ASPEN |
| Phase | Phase 3 |
| Condition | Waldenström's Macroglobulinemia |
| Design | Randomized, parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 201 |
| Arms | 2 |
| Interventions | BGB-3111; Ibrutinib |
| Status | Completed |
| Start | 2017-01-25 |
| Primary completion | 2022-06-21 |
| Lead sponsor | BeiGene |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT03053440 |
2. Clinical Question
The primary statistical question was whether the percentage of participants achieving either a complete response (CR) or very good partial response (VGPR) differed between the randomized BGB-3111 and ibrutinib groups during the registered follow-up period.
Population
Participants with Waldenström's macroglobulinemia enrolled in the phase 3 ASPEN trial.
Intervention
BGB-3111.
Comparator
Ibrutinib.
Primary question
Is there a difference between the treatment groups in the percentage of participants achieving CR or VGPR?
The registered hypothesis type was superiority. Thus, the primary analysis was structured to test whether the response percentage differed between the randomized groups rather than merely demonstrating that one treatment was not worse than the other.
3. Trial Design
Ibrutinib
- Ibrutinib was one of the two randomized interventions.
- The primary analysis compared this arm with the BGB-3111 arm.
BGB-3111 / Zanubrutinib
- BGB-3111 was the other randomized intervention.
- The registry-reported serious-adverse-event data identifies this arm as Zanubrutinib.
4. Endpoints
| Endpoint | Registry definition | Time frame | Type |
|---|---|---|---|
| Primary endpoint | Percentage of participants achieving either a Complete Response (CR) or Very Good Partial Response (VGPR) using an adaptation of the Response Criteria Updated at the Sixth International Workshop on WM as assessed by an Independent Review Committee. | Up to approximately 2 years and 7 months | Binary |
What counted as CR or VGPR?
The registered definition specifies CR as normal serum immunoglobulin M (IgM) levels, disappearance of monoclonal protein by immunofixation, and negative cryoglobulinemia if cryoglobulinemia was positive at baseline. VGPR was defined as a ≥90% reduction in serum IgM level from baseline or normal serum IgM values.
The endpoint therefore converts an underlying response assessment into a binary participant-level outcome: each participant is classified according to whether the prespecified CR-or-VGPR criterion was achieved.
5. Primary Result
Complete Response or Very Good Partial Response
Risk difference: Ibrutinib vs BGB-3111
95% CI: -1.5 to 22.0 · P = 0.0921
Analysis population: Intent To Treat (ITT) Analysis Set
| Primary endpoint | Analysis population | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|---|
| Percentage achieving CR or VGPR | Intent To Treat (ITT) Analysis Set | Cochran-Mantel-Haenszel | Risk Difference | 10.2 | -1.5 to 22.0 | 0.0921 |
The reported risk difference of 10.2 expresses an absolute difference in the percentage achieving CR or VGPR between the two randomized groups, using the comparison reported as Ibrutinib vs Zanubrutinib. In percentage-point terms, the estimated difference is 10.2 percentage points in the direction of the first-listed group in the analysis.
The estimate does not mean that an individual participant has a 10.2% higher probability of response, nor does it establish that the treatments differ by exactly 10.2 percentage points in the underlying population. It is a sample-based estimate subject to statistical uncertainty.
The 95% confidence interval of -1.5 to 22.0 shows the range of values compatible with the statistical estimation framework at the stated confidence level. Importantly, the interval crosses 0, the value representing no risk difference. That means the registry-reported interval does not exclude the possibility of no difference between the groups.
The p-value of 0.0921 is the two-sided probability measure reported for the Cochran-Mantel-Haenszel test under its null hypothesis framework. It does not measure the size of the treatment effect, the probability that either treatment is effective, or the probability that the null hypothesis is true. The effect estimate and confidence interval are needed to understand magnitude and precision.
Because this was a superiority analysis, the relevant comparison is whether the data provide sufficient evidence for a difference under the prespecified testing framework. The reported analysis was stratified according to the stratification factors per interactive response technology (IRT), so the result is not simply an unadjusted comparison of two raw percentages.
6. Statistical Methodology
Cochran-Mantel-Haenszel test
The registry reports the Cochran-Mantel-Haenszel test as the primary statistical method. This is a categorical-data method used when a comparison between two treatment groups is evaluated across one or more strata. Rather than treating the entire sample as a single unstructured 2 × 2 table, the method incorporates the stratified structure into the comparison.
The registry-reported analysis notes state that the test was stratified by the stratification factors per interactive response technology (IRT).
Risk difference
The reported effect measure was risk difference (RD). For a binary response endpoint, a risk difference compares the probability of the specified response between treatment groups on an absolute scale.
The reported ASPEN estimate is 10.2, with a 95% confidence interval from -1.5 to 22.0.
Risk difference is particularly useful when the scientific question concerns the absolute change in response probability. It differs from a relative risk or odds ratio because it is expressed directly in percentage-point terms rather than as a ratio.
Intention-to-treat analysis
The primary analysis used the Intent To Treat (ITT) Analysis Set. The ITT principle keeps participants associated with their randomized treatment assignment for the primary comparison. This preserves the treatment-group structure created by randomization and avoids redefining the primary comparison solely according to treatment actually received.
Stratified analysis
The analysis notes explicitly state that the Cochran-Mantel-Haenszel test was stratified by the stratification factors per interactive response technology (IRT). The ClinicalTrials.gov record does not identify those individual factors, so no specific strata are named here.
Superiority hypothesis
The registered hypothesis type was superiority. In a superiority framework, the question is whether the evidence supports a difference between the randomized treatment groups. This is conceptually different from non-inferiority, where the central question is whether a treatment remains within a prespecified acceptable loss relative to a comparator.
7. Statistical Methods Explained
Why use a Cochran-Mantel-Haenszel test?
The primary outcome is binary: participants either achieve CR or VGPR, or they do not. A Cochran-Mantel-Haenszel analysis is appropriate for comparing categorical outcomes while accounting for stratification. In ASPEN, the registry specifically reports that the test was stratified according to the stratification factors per IRT.
What does a risk difference of 10.2 mean?
A risk difference of 10.2 represents an estimated absolute separation of 10.2 percentage points between the groups in the probability of achieving the binary primary endpoint, with the direction determined by the reported group ordering. It is an absolute measure rather than a relative measure such as a risk ratio or odds ratio.
Why is the confidence interval important?
The confidence interval communicates uncertainty around the estimated risk difference. Here, the interval extends from -1.5 to 22.0. Because it includes 0, the interval includes both a small difference favoring the opposite direction and a larger positive difference under the reported comparison.
Why doesn't the p-value measure effect size?
The p-value addresses compatibility of the observed data with the null hypothesis used for the statistical test. It does not quantify how large the treatment difference is. Two studies can have similar effect estimates but different p-values because sample size and precision differ, and a statistically small p-value can occur with a small effect when precision is high.
Why does ITT matter for this endpoint?
Using the ITT Analysis Set means the primary treatment comparison remains tied to randomized assignment. This is important because randomization is the design feature that balances treatment assignment in expectation and provides the foundation for a causal comparison. Switching to a post-randomization subset can change the population being compared and potentially weaken that design advantage.
What does stratification add?
Stratification allows the analysis to account for the prespecified stratified structure rather than ignoring it. The Cochran-Mantel-Haenszel procedure combines information across strata into an overall comparison while respecting that structure. The registry does not provide the individual stratification variables in the ClinicalTrials.gov record, so their clinical interpretation cannot be reconstructed here.
Why is this not a hazard-ratio analysis?
The primary endpoint is a binary percentage of participants achieving CR or VGPR over the registered time frame. It is not defined as time to response, progression, or death. Therefore, the reported analysis uses categorical-data methodology and a risk difference rather than a Kaplan-Meier estimate or Cox proportional-hazards model.
8. Interpreting the Primary Endpoint in Absolute Terms
The primary effect measure is especially instructive because it is an absolute measure. If two treatments produce response probabilities that differ by 10.2 percentage points, the risk difference describes that separation directly. It does not require the reader to translate an odds ratio or relative risk into an absolute probability.
Absolute effect
Risk difference describes the difference in response probability directly, using percentage-point units.
Relative effect
A risk ratio or odds ratio would express the relationship as a ratio. Neither is the reported effect measure for this analysis.
Uncertainty
The 95% CI of -1.5 to 22.0 shows that the estimate is not precise enough to exclude 0 on the registry-reported two-sided interval.
Statistical evidence
The two-sided p-value of 0.0921 is the reported test result and should be interpreted together with the effect estimate and confidence interval.
This combination of an effect estimate, confidence interval, and p-value is more informative than the p-value alone. The estimate describes magnitude, the interval describes precision, and the p-value describes the result of the specified hypothesis test.
9. What the Primary Result Does — and Does Not — Establish
The reported risk difference of 10.2 is the estimated absolute difference between the two randomized groups for achieving CR or VGPR. It should not be interpreted as a universal treatment effect that applies identically to every patient.
The 95% CI of -1.5 to 22.0 indicates uncertainty around the estimated difference. Since 0 lies inside the interval, the data do not exclude no difference at the level represented by this two-sided interval.
The p-value of 0.0921 is a two-sided test result from the reported Cochran-Mantel-Haenszel analysis. It is not a probability that the treatment effect is 10.2, nor is it the probability that the null hypothesis is true.
The result concerns a specific composite response criterion: achievement of either CR or VGPR according to the registered response definitions. It should not automatically be generalized to other response categories or to endpoints that were not part of this analysis.
10. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized arm. The data identify the arms as Ibrutinib and Zanubrutinib for this safety measure.
| Safety measure | Ibrutinib | Zanubrutinib |
|---|---|---|
| Serious adverse events, affected / at risk | 50/98 | 59/101 |
These are reported counts of participants affected by serious adverse events divided by the number at risk in each arm. The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or p-value for serious adverse events.
The distinction between efficacy and safety populations also matters. The primary efficacy result is explicitly reported for the Intent To Treat (ITT) Analysis Set. The registry-reported serious-adverse-event information is presented by arm with affected/at-risk counts, but does not specify a separate named safety analysis population.
11. Planned Statistical Framework vs Reported Result
ASPEN illustrates a straightforward alignment between the endpoint type and the statistical method. The primary endpoint is binary, the effect measure is a risk difference, and the reported test is Cochran-Mantel-Haenszel with stratification.
| Statistical component | ASPEN | Interpretive role |
|---|---|---|
| Endpoint type | Binary | Participant-level CR/VGPR response classification |
| Analysis population | Intent To Treat (ITT) Analysis Set | Maintains randomized treatment assignment for primary efficacy analysis |
| Comparison | Ibrutinib vs Zanubrutinib | Randomized treatment-group comparison |
| Primary method | Cochran-Mantel-Haenszel | Stratified categorical-data comparison |
| Effect measure | Risk difference | Absolute difference in response probability |
| Estimate | 10.2 | Magnitude of the reported absolute difference |
| 95% CI | -1.5 to 22.0 | Precision and uncertainty around the estimate |
| P-value | 0.0921 | Two-sided result of the reported superiority test |
The analysis notes state that the Cochran-Mantel-Haenszel test was stratified by the stratification factors per IRT and that the p-value was two-sided. These details are important because the reported p-value is tied to the specified stratified analysis rather than to an unspecified unadjusted comparison.
12. Why Randomization Matters for the Analysis
Randomization is not merely a procedural feature; it determines how the statistical comparison can be interpreted. When participants are randomly assigned to treatment groups, the treatment assignment mechanism is separated from baseline prognostic factors in expectation. The primary ITT analysis then retains that randomized comparison.
Before treatment
Randomization establishes the treatment groups without selecting participants according to their eventual response.
At analysis
The ITT framework preserves the original randomized assignment for the primary efficacy comparison.
With stratification
The Cochran-Mantel-Haenszel procedure accounts for the trial's stratified analytical structure.
Interpretation
The statistical comparison concerns the randomized treatment strategies represented by the two arms, not an observational comparison of responders.
This is why it would be inappropriate to reinterpret the primary analysis simply as a comparison among participants who completed treatment or otherwise met a post-randomization condition when the registry explicitly identifies the ITT Analysis Set as the analysis population.
13. Confidence Intervals and Statistical Precision
The confidence interval is central to interpreting the ASPEN result. The point estimate alone is 10.2, but the interval extends from -1.5 to 22.0. That span is materially wider than the single point estimate and demonstrates why a point estimate should not be treated as if it were known without error.
The interval includes 0, the reference value corresponding to no absolute difference in response probability.
The interval also provides information about the direction and magnitude of uncertainty. Values below 0 correspond to a risk difference in the opposite direction, whereas positive values correspond to a difference in the direction of the reported first-listed treatment comparison. The interval therefore does more than label the result "significant" or "nonsignificant": it describes the range of effect sizes that remain compatible with the statistical uncertainty represented by the analysis.
14. P-values in Context
The reported 0.0921 p-value is two-sided. A two-sided test evaluates departures from the null in either direction rather than assigning all of the rejection region to one direction.
What it tells us
It summarizes the result of the specified two-sided Cochran-Mantel-Haenszel hypothesis test under its null framework.
What it does not tell us
It does not give the probability that the treatment effect is absent, nor does it quantify clinical importance.
Read with the estimate
The p-value should be considered alongside the risk difference of 10.2 and its 95% CI of -1.5 to 22.0.
Read with the design
The reported result comes from an ITT, stratified analysis rather than an unspecified unadjusted comparison.
A useful statistical reading therefore avoids replacing the entire result with a single p-value. The effect estimate describes the observed difference, while the confidence interval and hypothesis test communicate uncertainty and evidence under the prespecified analytical framework.
15. Serious Adverse Events: Statistical Perspective
The ClinicalTrials.gov record provides serious-adverse-event counts of 50/98 for Ibrutinib and 59/101 for Zanubrutinib. These figures are descriptive safety information.
| Question | What the ClinicalTrials.gov record supports | What they do not support |
|---|---|---|
| How many were affected? | 50/98 in Ibrutinib and 59/101 in Zanubrutinib. | No additional event-level breakdown is reported. |
| Is there a formal between-arm test? | No formal statistical analysis is posted on ClinicalTrials.gov for this safety measure. | |
| Is this the primary efficacy population? | The primary efficacy analysis is explicitly identified as ITT. | The ClinicalTrials.gov record does not explicitly name a separate safety analysis set. |
This is a useful reminder that a numerical safety table and a formal comparative statistical analysis are not the same thing. Descriptive counts can communicate the observed safety experience without establishing a hypothesis-tested treatment difference.
16. Missing Data, Crossover, and Interim Analysis
The ClinicalTrials.gov record does not report a specific missing-data or imputation method for the primary endpoint. They also do not report a crossover strategy or an interim-analysis procedure.
For a binary response endpoint, missing response assessments can matter because the denominator and classification of participants affect the estimated response percentage. The appropriate handling depends on the prespecified analysis plan and estimand. Without a reported imputation rule or missing-data strategy in the ClinicalTrials.gov record, no such procedure should be attributed to ASPEN.
Similarly, interim analyses can affect type I error control if repeated formal testing is performed before the final analysis. Because the ClinicalTrials.gov record does not report an interim-analysis framework, no alpha-spending or group-sequential method is assigned to this trial on this page.
17. Multiplicity and Endpoint Interpretation
The ClinicalTrials.gov record identifies 1 registered primary endpoint and report 14 outcome measures overall, but only 1 statistical analysis is in the ClinicalTrials.gov record. The primary endpoint is the CR-or-VGPR percentage.
| Measure | Registry information reported | Interpretation |
|---|---|---|
| Registered primary endpoints | 1 | One endpoint is identified as primary. |
| Outcome measures posted | 14 | The registry contains additional outcome measures beyond the primary endpoint. |
| Statistical analyses posted | 1 | One formal statistical analysis is reported in the ClinicalTrials.gov record. |
| Primary analyses with estimate + CI | 1 | The registry-reported primary analysis includes an estimate and confidence interval. |
The existence of multiple outcome measures does not by itself establish a multiplicity adjustment strategy. To claim control of familywise type I error across multiple endpoints or analyses, the prespecified testing hierarchy or multiplicity procedure would need to be documented. That information is not included in the ClinicalTrials.gov record.
18. Trial Timeline
Trial start
The ASPEN phase 3 trial began on January 25, 2017.
Parallel treatment comparison
Participants were randomized in a parallel design to BGB-3111 or ibrutinib, with no masking.
CR or VGPR
The primary endpoint assessed the percentage of participants achieving either CR or VGPR during the registered time frame of up to approximately 2 years and 7 months.
Primary completion
The trial reached primary completion on June 21, 2022.
Registry results posted
Results were posted, including the formal primary-endpoint statistical analysis reported on this page.
19. Why This Trial Matters Statistically
ASPEN is a useful teaching example because its primary analysis connects a clinically meaningful binary response endpoint to a stratified categorical-data method and an absolute effect measure. The result can be understood without relying on a hazard ratio or time-to-event model.
| Concept | How it appears in ASPEN |
|---|---|
| Randomization | The phase 3 trial used randomized allocation to 2 treatment arms. |
| Parallel design | The trial is registered with a parallel design model. |
| Binary endpoint | CR or VGPR achievement is analyzed as a binary participant-level outcome. |
| ITT analysis | The primary analysis used the Intent To Treat (ITT) Analysis Set. |
| Cochran-Mantel-Haenszel test | The primary comparison used a stratified Cochran-Mantel-Haenszel analysis. |
| Risk difference | The treatment effect was reported as an absolute risk difference. |
| Confidence interval | The 95% CI was -1.5 to 22.0. |
| P-value | The reported two-sided p-value was 0.0921. |
| Superiority testing | The registered hypothesis type was superiority. |
| Stratified analysis | The analysis was stratified by the stratification factors per IRT. |
| Safety reporting | Serious adverse events were reported as affected/at-risk counts by arm. |
The broader statistical lesson is that the analysis should follow the endpoint. A binary response endpoint does not require a survival model merely because the trial itself has a follow-up period. The relevant distinction is between when the endpoint is assessed and what type of outcome is being analyzed.
20. Limitations and Interpretation Issues
- Limited formal results in the ClinicalTrials.gov record: only one statistical analysis is reported, corresponding to the primary endpoint. Other posted outcome measures are not accompanied here by formal comparative estimates.
- Confidence interval crosses zero: the 95% CI of -1.5 to 22.0 includes the no-difference value for a risk difference.
- Binary endpoint specificity: the result applies to the registered CR-or-VGPR definition and should not automatically be generalized to other response definitions.
- Stratification details are incomplete: the analysis states that stratification was by factors per IRT, but the individual factors are not reported.
- Missing-data methods are not reported: the ClinicalTrials.gov record does not specify how missing response assessments were handled.
- No interim-analysis details are reported: the ClinicalTrials.gov record does not establish whether formal interim testing or alpha spending was used.
- No crossover information is reported: no crossover procedure is described in the ClinicalTrials.gov record, so its potential influence cannot be quantified here.
- Safety analysis is descriptive: serious-adverse-event counts are provided, but no formal comparative test is reported.
- Multiplicity cannot be reconstructed: the registry contains 14 outcome measures, but the ClinicalTrials.gov record does not specify a multiplicity-control strategy across them.
- Open-label design: masking is listed as none. For a response endpoint, the use of an independent review committee is an important part of the registered endpoint definition, but the ClinicalTrials.gov record does not provide additional details about the review process.
21. A Structured Reading of the Primary Result
Step 1 — Identify the endpoint
Determine whether a participant achieved CR or VGPR according to the registered response definition.
Step 2 — Identify the population
The formal analysis uses the Intent To Treat Analysis Set.
Step 3 — Identify the method
The reported method is a Cochran-Mantel-Haenszel test stratified according to the IRT stratification factors.
Step 4 — Identify the effect
The effect measure is a risk difference of 10.2.
Step 5 — Assess precision
The 95% confidence interval ranges from -1.5 to 22.0.
Step 6 — Assess the test result
The reported two-sided p-value is 0.0921.
This sequence prevents a common analytical mistake: starting with the p-value and attempting to infer the entire statistical story from it. The more complete interpretation starts with the endpoint, population, analysis method, effect measure, confidence interval, and then the hypothesis-test result.
22. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary analysis estimated a risk difference of 10.2, with a 95% CI of -1.5 to 22.0 and a two-sided p-value of 0.0921 using a stratified Cochran-Mantel-Haenszel test in the ITT Analysis Set.
Clinical interpretation
The endpoint concerns achievement of CR or VGPR under a specific response definition. The ClinicalTrials.gov record alone do not justify translating the statistical result into treatment recommendations or a broader clinical judgment.
Keeping these interpretations separate is important. Statistical evidence describes the strength and uncertainty of the observed treatment-group comparison. Clinical interpretation additionally depends on the endpoint's importance, durability, safety, patient population, treatment context, and other evidence. Only the statistical evidence supported by the ClinicalTrials.gov record is quantified here.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: NCT03053440 — ASPEN.
- PubMed: PMID 29869556.
- PubMed: PMID 40924923.
- PubMed: PMID 39626287.
- PubMed: PMID 39072392.
- PubMed: PMID 38502198.
Continue through Clinical Biostats
Explore statistical tutorials and calculators connected to the methods used in randomized clinical trials.
26. Record Summary
ASPEN provides a clear example of how a binary clinical-trial endpoint can be analyzed using a stratified categorical-data method. The primary endpoint measured the percentage of participants achieving either CR or VGPR according to the registered response definition. The formal analysis used the Intent To Treat Analysis Set and a Cochran-Mantel-Haenszel test stratified by the stratification factors per IRT, with risk difference as the effect measure.
The reported estimate was 10.2, with a 95% confidence interval of -1.5 to 22.0 and a two-sided p-value of 0.0921. The most informative reading therefore combines the absolute effect estimate, its uncertainty interval, the hypothesis-test result, and the analytical structure that produced them. The ClinicalTrials.gov record add descriptive serious-adverse-event counts of 50/98 for Ibrutinib and 59/101 for Zanubrutinib, but no formal comparative safety analysis is provided in the ClinicalTrials.gov recordset.