This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Reported estimates are kept separate from the interpretive commentary, which is labelled as such.
1. Trial at a Glance
MANIFEST-2 was a randomized, placebo-controlled phase 3 trial asking whether adding pelabresib, an investigational BET inhibitor, to the JAK inhibitor ruxolitinib increases the proportion of patients with myelofibrosis who achieve a spleen volume reduction of at least 35% (SVR35) at week 24.
| Feature | MANIFEST-2 |
|---|---|
| Phase | Phase 3 |
| Conditions | Myelofibrosis; primary myelofibrosis; post-polycythemia vera myelofibrosis; post-essential thrombocythemia myelofibrosis |
| Design | Randomized, parallel-group, quadruple-masked, placebo-controlled, two arms |
| Enrollment | 430 |
| Primary endpoint | Splenic response (SVR35) by blinded central radiology review at week 24 |
| Key secondary endpoints | Absolute change from baseline in Total Symptom Score (TSS) at week 24; TSS50 response at week 24 |
| Primary analysis | Cochran-Mantel-Haenszel (CMH) comparison, difference in proportions with stratum-adjusted 95% CI, ITT analysis set |
| Status | Active, not recruiting (start April 9, 2021; primary completion August 23, 2023) |
| ClinicalTrials.gov | NCT04603495 |
| Lead sponsor | Novartis Pharmaceuticals |
2. Clinical Question
Ruxolitinib reduces spleen size and symptom burden in many patients with myelofibrosis, but responses are incomplete. MANIFEST-2 tested an add-on strategy: every participant received ruxolitinib, and randomization determined whether pelabresib or placebo was added. The contrast therefore isolates the incremental effect of pelabresib on top of a common active backbone.
Population
Patients with myelofibrosis, including primary myelofibrosis, post-polycythemia vera myelofibrosis and post-essential thrombocythemia myelofibrosis.
Intervention
Pelabresib (CPI-0610) plus ruxolitinib.
Comparator
Placebo plus ruxolitinib.
Primary question
Does adding pelabresib to ruxolitinib increase the proportion of participants achieving at least a 35% reduction in spleen volume from baseline at week 24, as assessed by blinded central radiology?
3. Trial Design
Pelabresib + Ruxolitinib
- Pelabresib (CPI-0610)
- Ruxolitinib
Placebo + Ruxolitinib
- Placebo matched to pelabresib
- Ruxolitinib
4. Stratification and Analysis Populations
The posted analyses describe stratum-adjusted CMH estimates, with strata including baseline DIPSS category, platelet count and spleen volume. Because few participants fell into the DIPSS High category, it was merged with Intermediate-2 for analysis, a practical step that avoids strata too small to contribute stable information.
| Analysis population | Used for |
|---|---|
| Intent-to-Treat (ITT) analysis set | Primary SVR35 analysis; TSS change; TSS50 response; mTSS response |
| ITT analysis set, participants with available data | Bone marrow fibrosis improvement at week 24 (only participants with data at that time point) |
| On-treatment safety population | Serious adverse events: 212 at risk with pelabresib + ruxolitinib; 214 with placebo + ruxolitinib |
For binary response endpoints analyzed in the ITT set, participants without a week-24 assessment would normally be counted as non-responders, which keeps the full randomized denominator. The bone marrow fibrosis analysis instead used observed data only, so it does not carry the same protection: if the reasons for missing biopsies differed between arms, the comparison could be biased in either direction.
5. Endpoints
Primary endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Number of participants with splenic response by central radiology reads at week 24 | Splenic response characterized by a reduction of at least 35% in spleen volume from baseline (SVR35), determined by MRI or CT and evaluated through a blinded central radiology review at week 24. | Week 24 |
Secondary endpoints with posted analyses
| Endpoint | Time frame | Method / effect measure |
|---|---|---|
| Key secondary: absolute change from baseline in Total Symptom Score (TSS) | Baseline, week 24 | ANCOVA; least-squares (LS) mean difference |
| Key secondary: number of participants with TSS50 response | Week 24 | CMH; difference in proportions |
| Improvement from baseline in bone marrow fibrosis of at least 1 grade | Baseline, week 24 | CMH; difference in proportions |
| Modified Total Symptom Score (mTSS) response | Week 24 | CMH; difference in proportions (Wald 95% CI) |
6. Primary Endpoint Result: SVR35 at Week 24
The registry reports the difference in the proportion of participants achieving SVR35, computed as experimental arm (pelabresib + ruxolitinib) minus control arm (placebo + ruxolitinib), with a CMH 95% confidence interval adjusted by strata in the ITT analysis set.
Difference in SVR35 response rate (percentage points)
95% CI: 21.6 to 39.3 (two-sided) · P < 0.001
Pelabresib + ruxolitinib minus placebo + ruxolitinib · Cochran-Mantel-Haenszel · ITT analysis set
| Quantity | Reported value |
|---|---|
| Effect measure | Difference in proportion (experimental − control) |
| Estimate | 30.4 |
| 95% confidence interval (two-sided) | 21.6 to 39.3 |
| P-value | <0.001 |
| Method | CMH test, CI adjusted by strata; Mehrotra-Railkar test used if the Breslow-Day test was significant; DIPSS High merged with Int-2 |
| Analysis population | Intent-to-Treat (ITT) analysis set |
What the estimate means. A difference of 30.4 means that, after combining information across strata, the estimated proportion of participants achieving SVR35 at week 24 was 30.4 percentage points higher with pelabresib + ruxolitinib than with placebo + ruxolitinib. This is an absolute risk difference, which translates directly into how many additional responders one would expect per 100 patients treated under trial conditions.
What it does not mean. It is not a relative increase of 30.4%, and it does not describe how much any individual's spleen shrank. SVR35 is a dichotomy: a participant with a 34% reduction and one with no reduction are both non-responders. The difference also says nothing on its own about symptoms, survival or disease modification.
Precision. The 95% CI of 21.6 to 39.3 percentage points is well away from zero. Even its lower limit corresponds to a substantial absolute difference, so the direction and approximate size of the effect on this endpoint are well determined.
The P-value. P < 0.001 indicates the observed difference would be very unlikely if pelabresib added nothing to response. The P-value itself does not convey the magnitude; the 30.4-point estimate and its interval do.
Cautions. The CMH estimate is a weighted average of stratum-specific differences and assumes a broadly common difference across strata; the prespecified switch to a Mehrotra-Railkar approach if Breslow-Day indicated heterogeneity addresses this. The endpoint is measured at a single time point, so it does not describe durability of response beyond week 24.
7. Key Secondary Results: Symptom Burden
The two key secondary endpoints address whether spleen responses were accompanied by symptom improvement.
| Endpoint | Effect measure | Estimate (95% CI) | P-value |
|---|---|---|---|
| Absolute change from baseline in TSS, week 24 | LS mean difference (ANCOVA) | −1.94 (−3.92 to 0.04) | 0.0545 |
| TSS50 response, week 24 | Difference in proportions (CMH) | 6.0 (−3.4 to 15.5) | 0.216 |
TSS change. The LS mean difference of −1.94 points indicates a numerically larger mean reduction in symptom score in the pelabresib arm, after ANCOVA adjustment. The 95% CI (−3.92 to 0.04) just includes zero and P = 0.0545 lies just above 0.05. The data are compatible with a benefit of nearly 4 points and with essentially no difference. The registry does not report a minimal clinically important difference against which the size of −1.94 could be judged.
TSS50. The proportion achieving at least a 50% symptom reduction was 6.0 percentage points higher with pelabresib, but the interval (−3.4 to 15.5) is wide and includes both a small deficit and a moderate benefit.
A near-miss is not a hidden success. When a key secondary endpoint sits in a formal testing sequence, a P-value of 0.0545 against a 0.05 threshold is not significant, and later endpoints in the same sequence cannot be claimed as confirmatory. It is equally wrong to read the result as showing "no effect on symptoms"; the interval is what should be reported.
8. Other Secondary Results: Bone Marrow Fibrosis and mTSS
| Endpoint | Analysis population | Difference in proportions (95% CI) | P-value |
|---|---|---|---|
| ≥1-grade improvement in bone marrow fibrosis, week 24 | ITT, participants with available data | 7.63 (0.52 to 14.73) | 0.037 |
| mTSS response, week 24 | ITT | 7.8 (−1.7 to 17.2) | 0.107 |
The bone marrow fibrosis result is nominally significant, with a lower confidence limit close to zero (0.52 percentage points). Given that this analysis was restricted to participants with an available week-24 biopsy, that its role in any testing hierarchy is not described in the registry, and that its interval barely excludes zero, it is best read as hypothesis-generating. The mTSS response difference of 7.8 points has an interval including zero; the registry notes that Wald 95% CIs were applied for this analysis.
Differences in proportions at a glance
9. Safety: Serious Adverse Events
| Safety measure | Pelabresib + Ruxolitinib (on-treatment) | Placebo + Ruxolitinib (on-treatment) |
|---|---|---|
| Participants with serious adverse events (affected / at risk) | 63 / 212 | 63 / 214 |
The same number of participants (63) had a serious adverse event in each arm, against nearly identical denominators. No formal statistical comparison of serious adverse events is posted. A count of participants with any SAE is a coarse summary: it does not distinguish event types, severity, timing or relationship to treatment, and similar totals can conceal differences in specific events. It should be read as one component of the safety evidence, separate from the efficacy endpoints.
10. Statistical Methodology
Cochran-Mantel-Haenszel analysis of response proportions
The primary endpoint and three secondary response endpoints were analyzed with the CMH approach. Rather than pooling all participants into one 2×2 table, the CMH method computes the treatment difference within each stratum and combines those differences with weights, so that the comparison respects the stratified design.
where pE,k and pC,k are the response proportions in the experimental and control arms within stratum k. Strata with more balanced and larger sample sizes receive more weight.
Breslow-Day test and the Mehrotra-Railkar alternative
The CMH summary is most interpretable when the treatment effect is similar across strata. The analysis plan used the Breslow-Day test to check for heterogeneity across strata and specified a switch to the Mehrotra-Railkar test if heterogeneity was significant. The Mehrotra-Railkar approach uses "minimum risk" weights designed to remain efficient when stratum-specific differences vary, making the overall comparison less dependent on the common-effect assumption.
ANCOVA for change in Total Symptom Score
The absolute change from baseline in TSS at week 24 was analyzed by analysis of covariance, and the effect is reported as a least-squares (LS) mean difference between arms.
βtrt is the LS mean difference, reported as −1.94. The registry does not list the model covariates; baseline score and stratification factors are the conventional choices for this type of analysis. Adjusting for baseline reduces residual variance and guards against chance baseline imbalance.
Wald confidence intervals
For mTSS response, the registry notes that Wald 95% CIs were applied: estimate ± 1.96 × standard error. Wald intervals are simple and generally adequate at the sample sizes of this trial, although they can be less accurate when proportions are near 0 or 1.
Intention-to-treat principle
The primary analysis used the ITT analysis set, so participants were analyzed according to their randomized arm. This preserves the comparability created by randomization even if some participants discontinued pelabresib or placebo before week 24.
11. Multiplicity and Endpoint Hierarchy
MANIFEST-2 had one primary endpoint and two endpoints labelled in the registry as key secondary. A "key secondary" designation usually indicates endpoints placed within a prespecified, multiplicity-controlled testing strategy, but the registry does not describe the testing order or alpha allocation.
| Analysis | Role | Interpretation |
|---|---|---|
| SVR35 at week 24 | Primary | Confirmatory comparison under the prespecified design |
| TSS change; TSS50 response | Key secondary | Tested after the primary; neither reached P < 0.05 |
| Bone marrow fibrosis improvement; mTSS response | Secondary | Supportive; nominal P-values |
| SAE counts | Safety | Descriptive; no formal test posted |
12. Statistical Methods Explained
Why was a CMH test used rather than a simple chi-squared test?
The analysis was stratified by prognostic factors (DIPSS category, platelet count and spleen volume). The CMH method compares arms within each stratum and then combines the results, matching the analysis to the design. This typically improves precision and avoids distortions that can arise when strata with different baseline response rates are pooled into a single table.
What does a difference in proportions of 30.4 mean?
It is an absolute difference: roughly 30 more participants per 100 achieved SVR35 at week 24 with pelabresib added than with placebo added. An absolute difference is often easier to relate to clinical practice than a relative measure, because it directly reflects the number of additional responders. Its reciprocal is the number needed to treat for one additional SVR35 response under trial conditions.
Why was DIPSS High merged with Intermediate-2?
Stratified estimators need enough participants, and enough responders and non-responders, in each stratum. A stratum with very few participants can produce unstable or undefined stratum-specific estimates. Merging the sparse High category with the adjacent Intermediate-2 category keeps the stratified analysis estimable while preserving most of the prognostic information.
Why was an ANCOVA used for symptom score change?
TSS change is a continuous outcome, and people with higher baseline symptoms have more room to improve. ANCOVA adjusts the comparison for baseline and other covariates, which reduces unexplained variability and produces a more precise estimate than comparing raw mean changes. The result is expressed as an LS mean difference, the adjusted between-arm difference in mean change.
How should P = 0.0545 for TSS change be read?
It is slightly above the conventional 0.05 threshold and the 95% CI (−3.92 to 0.04) just crosses zero. Under a prespecified hypothesis test, the endpoint did not meet statistical significance. The interval shows that the data are compatible with a modest symptom benefit and with no difference, so the appropriate conclusion is uncertainty, not equivalence.
Why can a large spleen response coexist with an inconclusive symptom result?
Both arms received ruxolitinib, which itself improves symptoms, so the room for an additional symptom effect may be smaller than for spleen volume. Symptom scores are also more variable than imaging measurements, which widens confidence intervals. A clear effect on one endpoint and an imprecise estimate on another are statistically compatible and do not contradict each other.
13. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary SVR35 comparison produced a 30.4-percentage-point difference with a 95% CI well above zero. Neither key secondary symptom endpoint reached P < 0.05, and their intervals include zero.
Clinical interpretation
Adding pelabresib to ruxolitinib increased the proportion of patients with a substantial spleen volume reduction at week 24. Whether this translates into meaningful symptom relief was not established at the confirmatory level, and longer-term outcomes are not part of the registry's posted analyses.
14. Important Limitations and Interpretation Issues
- Single time point: the primary endpoint is assessed at week 24 and does not describe durability of response.
- Dichotomized endpoint: SVR35 discards information about the size of spleen volume change; participants near the 35% threshold are classified sharply.
- Surrogate nature: spleen volume reduction is a widely used endpoint in myelofibrosis trials but is not itself a measure of survival or disease transformation.
- Key secondary endpoints not met: TSS change (P = 0.0545) and TSS50 (P = 0.216) did not reach significance, which limits claims about symptom benefit.
- Observed-data analysis: the bone marrow fibrosis comparison included only participants with available data and may be affected by missing biopsies.
- Multiplicity: the registry does not describe the full testing hierarchy, so nominal P-values for non-key secondary endpoints should be treated as exploratory.
- Unreported model details: ANCOVA covariates and handling of missing symptom data are not described in the registry.
- Ongoing study: the trial is active, not recruiting; additional follow-up may alter the long-term picture.
15. Why This Trial Matters Statistically
MANIFEST-2 is a clear example of how binary response endpoints, stratified analysis and continuous patient-reported outcomes are combined in a modern add-on trial.
| Concept | How it appears in MANIFEST-2 |
|---|---|
| Add-on placebo-controlled design | Both arms receive ruxolitinib; randomization isolates the effect of pelabresib |
| Risk difference | Primary effect expressed as a 30.4-percentage-point difference in SVR35 |
| Cochran-Mantel-Haenszel test | Stratum-adjusted comparison of response proportions |
| Heterogeneity across strata | Breslow-Day test with prespecified Mehrotra-Railkar alternative |
| Sparse strata | DIPSS High merged with Int-2 |
| ANCOVA | Baseline-adjusted LS mean difference in TSS change |
| Confidence intervals | Separate a well-determined primary effect from imprecise symptom effects |
| Intention-to-treat | Primary and key secondary analyses in the ITT analysis set |
| Multiplicity | Primary, key secondary and other secondary endpoints carry different evidential weight |
16. Related Tutorials
Learn more about the methods used in this trial:
17. Related Calculators
18. Sources
- ClinicalTrials.gov registry record: NCT04603495 — Phase 3 Study of Pelabresib (CPI-0610) in Myelofibrosis (MANIFEST-2).
- PubMed: PMID 35950489.
- PubMed: PMID 40065169.
Explore the methods behind the results
Follow the endpoints and analyses used in this trial to tutorials and calculators that explain how they work.
19. Record Summary
MANIFEST-2 enrolled 430 participants with myelofibrosis and compared pelabresib + ruxolitinib with placebo + ruxolitinib. The stratified CMH analysis of the primary endpoint in the ITT analysis set showed a 30.4-percentage-point higher SVR35 rate at week 24 (95% CI 21.6 to 39.3; P < 0.001). The key secondary symptom endpoints did not reach significance: the TSS LS mean difference was −1.94 (95% CI −3.92 to 0.04; P = 0.0545) and the TSS50 difference was 6.0 (95% CI −3.4 to 15.5; P = 0.216). Serious adverse events occurred in 63 participants in each arm. The most useful reading combines the absolute effect on spleen response, the stratified design behind it, the uncertainty around symptom effects, and the testing hierarchy that determines which results are confirmatory.