← Clinical Trials
Chronic Migraine Prevention Phase 3 Randomized NCT02974153

PROMISE-2: Complete Statistical Analysis of Eptinezumab in Chronic Migraine

An independent statistical review of the randomized phase 3 PROMISE-2 trial evaluating ALD403 (eptinezumab) in the prevention of chronic migraine, with emphasis on its continuous and binary efficacy endpoints and the ANCOVA, Cochran-Mantel-Haenszel, and repeated-measures methods reported in the registry.

PROMISE-2  ·  Phase 3  ·  Completed  ·  Enrollment 1121
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results are restricted to the information posted for NCT02974153 in the ClinicalTrials.gov record. The registry record is the official source for the trial design and posted statistical analyses.

Registry note: 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

PROMISE-2 was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating ALD403 (eptinezumab) for the prevention of chronic migraine in participants with migraine disorders. The trial enrolled 1121 participants across three arms and reported efficacy analyses comparing 300 mg ALD403 and 100 mg ALD403 with placebo.

1121
Enrollment
Participants
3
Arms
ALD403 and placebo
3
Phase
Phase 3
16
Analyses posted
2 primary; 14 secondary
FeaturePROMISE-2
Trial namePROMISE-2
PhasePhase 3
ConditionMigraine Disorders
Brief titleEvaluation of ALD403 (Eptinezumab) in the Prevention of Chronic Migraine
AllocationRandomized
Design modelParallel
MaskingQuadruple
Primary purposePrevention
Enrollment1121
Arms3
Lead sponsorAlder Biopharmaceuticals, Inc.
Sponsor typeIndustry
StatusCompleted
Start2016-11
Primary completion2017-11
ClinicalTrials.govNCT02974153

2. Clinical Question

The clinical question was whether ALD403 (eptinezumab), administered at either 300 mg or 100 mg, produced a greater improvement than placebo in the registered primary outcome, change from baseline in monthly migraine days, during Weeks 1-12.

Population

Participants enrolled in a phase 3 prevention trial for chronic migraine, with the registered condition listed as Migraine Disorders.

Intervention

ALD403 (eptinezumab), evaluated at 300 mg and 100 mg in the reported treatment comparisons.

Comparator

Placebo.

Primary question

Does ALD403 produce a greater reduction in monthly migraine days than placebo over Weeks 1-12?

3. Trial Design

01
Randomize1121 enrolled
02
3 ArmsTwo ALD403 doses + placebo
03
MaskQuadruple-masked
04
AssessMonthly migraine outcomes
05
CompareALD403 vs placebo
Allocation
Randomized allocation in a parallel-group design.
Masking
Quadruple masking.
Primary purpose
Prevention.
Statistical profile
ANCOVA for continuous outcomes, Cochran-Mantel-Haenszel testing for binary outcomes, and MMRM for a repeated-measures endpoint.
ARM 1

300 mg ALD403

  • ALD403 (eptinezumab)
  • Reported as a randomized treatment arm
  • Compared with placebo in the posted statistical analyses
ARM 2

100 mg ALD403

  • ALD403 (eptinezumab)
  • Reported as a randomized treatment arm
  • Compared with placebo in the posted statistical analyses
ARM 3

Placebo

  • Placebo
  • Comparator arm for the reported efficacy analyses
Analysis population: The posted statistical analyses use the Full Analysis Population, defined in the ClinicalTrials.gov record as all randomized participants who received investigational product or placebo. This population definition is used for both the primary and secondary efficacy analyses reported here.

4. Endpoints

The registry lists one primary endpoint and reports 27 outcome measures overall. The registered primary endpoint is a continuous measure based on monthly migraine days.

Endpoint roleRegistered endpointTime frameRegistry definition
Primary Change From Baseline in Monthly Migraine Days Week 1-12 Monthly migraine days are summarized in 28-day intervals, and averaged across Weeks 1-12.

Secondary endpoints with posted statistical analyses

Outcome measureTime frameEndpoint typeReported method
75% Migraine Responder RateWeek 1-12BinaryCochran-Mantel-Haenszel
75% Migraine Responder Rate - 4 WeekWeek 1-4BinaryCochran-Mantel-Haenszel
50% Migraine Responder RateWeek 1-12BinaryCochran-Mantel-Haenszel
Change in Monthly Acute Medication DaysWeek 1-12ContinuousANCOVA
Change From Baseline of Headache Impact Test (HIT-6) ScoreBaseline to Week 12ContinuousANCOVA
Change in Migraine Prevalence From Baseline to Week 4Baseline to Week 4BinaryMMRM (reported as Repeated Measures Model)
Percentage of Participants With a Migraine on the Day After DosingDay 1BinaryCochran-Mantel-Haenszel

5. Statistical Methodology

ANCOVA for continuous outcomes

The primary endpoint, change from baseline in monthly migraine days, was analyzed using analysis of covariance (ANCOVA). The same method was used for change in monthly acute medication days and change from baseline in HIT-6 score.

ANCOVA is a linear-model approach that compares treatment groups while accounting for baseline information and other prespecified covariates when included in the model. For a change-from-baseline endpoint, this is particularly useful because baseline outcome levels can be related to the subsequent measurement.

Conceptual ANCOVA structure
Yi = β0 + β1Treatmenti + β2Baselinei + εi

Here, the treatment coefficient represents an adjusted between-group difference in the modeled outcome, conditional on the covariates included in the actual analysis.

Cochran-Mantel-Haenszel testing

The 75% and 50% migraine responder outcomes, as well as the Day 1 migraine outcome, were analyzed with the Cochran-Mantel-Haenszel method. This is a categorical-data framework that can compare treatment groups while accounting for stratification variables when such strata are part of the analysis.

The registry's reported effect measure for these analyses is Mean Difference (Final Values). The ClinicalTrials.gov record does not provide a risk ratio, odds ratio, or the underlying arm-specific responder proportions for these analyses. Accordingly, the results below retain the registry's reported effect-measure terminology rather than converting them into a different measure.

MMRM for repeated measurements

The change in migraine prevalence from baseline to Week 4 was analyzed using a Repeated Measures Model, normalized in the ClinicalTrials.gov record as an MMRM, or mixed model for repeated measures.

An MMRM is designed for longitudinal data in which the same participant contributes measurements at multiple time points. Rather than treating each time point as an entirely separate analysis, the model can represent the within-participant correlation across repeated observations.

Conceptual longitudinal structure
Yij = fixed effects + participant-level repeated-measure structure + errorij

The actual covariance structure and full model specification are not provided in the ClinicalTrials.gov record, so this page does not infer them.

Superiority framework

All 16 posted statistical analyses are identified as superiority comparisons. Thus, the treatment comparisons are framed as testing whether the ALD403 group differs favorably from placebo rather than as non-inferiority or equivalence assessments.

Two-sided confidence intervals

The posted estimates with confidence intervals use 95% two-sided confidence intervals. A confidence interval communicates the statistical precision of the estimated treatment difference; it is not a range containing the effects experienced by individual participants.

6. Results: Primary Endpoint

The registry reports two formal primary-endpoint analyses for the same registered outcome: one comparing 300 mg ALD403 with placebo and one comparing 100 mg ALD403 with placebo. Both use ANCOVA in the Full Analysis Population.

300 mg ALD403 vs Placebo

Mean difference in change from baseline

-2.60

95% CI: -3.45 to -1.74   ·   P < 0.0001

Outcome: Change From Baseline in Monthly Migraine Days, averaged across Weeks 1-12.

FeatureReported result
EndpointChange From Baseline in Monthly Migraine Days
Time frameWeek 1-12
Comparison300 mg ALD403 vs Placebo
PopulationFull Analysis Population - all randomized participants who received investigational product or placebo
MethodANCOVA
Effect measureMean Difference (Final Values)
Estimate-2.60
95% CI-3.45 to -1.74
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

The estimate of -2.60 is the reported adjusted mean difference for the change-from-baseline outcome, comparing 300 mg ALD403 with placebo. Because the outcome is change in monthly migraine days, the negative direction indicates a more negative modeled change for the ALD403 group than for placebo.

The estimate does not mean that every participant experienced exactly 2.60 fewer migraine days. It is a group-level statistical estimate, not an individual treatment effect.

The 95% CI of -3.45 to -1.74 describes uncertainty around the estimated between-group difference under the analysis framework. Its width gives a sense of statistical precision, but it does not describe the range of individual responses.

The P < 0.0001 value addresses the evidence against the null hypothesis specified by the superiority comparison. It does not measure the size, clinical importance, or probability of the treatment effect.

Because this is an ANCOVA result rather than a time-to-event analysis, proportional-hazards assumptions are not relevant to this estimate. Interpretation should instead remain tied to the modeled continuous endpoint, the Full Analysis Population, and the prespecified analysis model.

100 mg ALD403 vs Placebo

Mean difference in change from baseline

-2.03

95% CI: -2.88 to -1.18   ·   P < 0.0001

Outcome: Change From Baseline in Monthly Migraine Days, averaged across Weeks 1-12.

FeatureReported result
EndpointChange From Baseline in Monthly Migraine Days
Time frameWeek 1-12
Comparison100 mg ALD403 vs Placebo
PopulationFull Analysis Population - all randomized participants who received investigational product or placebo
MethodANCOVA
Effect measureMean Difference (Final Values)
Estimate-2.03
95% CI-2.88 to -1.18
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

The reported estimate of -2.03 represents the modeled difference in change from baseline in monthly migraine days between 100 mg ALD403 and placebo over Weeks 1-12. The negative estimate indicates a more negative change in the ALD403 group relative to placebo.

It does not mean that each treated participant had exactly 2.03 fewer migraine days, nor does it establish that every participant benefited to the same degree.

The 95% CI of -2.88 to -1.18 provides the statistical uncertainty associated with the estimated group difference. A narrower interval would represent greater statistical precision, but the interval itself does not describe individual-level variability.

The P < 0.0001 result provides evidence against the relevant null hypothesis under the posted superiority analysis. A p-value is not an effect-size measure and should not be interpreted as the probability that the treatment works.

The two primary comparisons also raise the issue of multiplicity: two active-dose comparisons are reported against the same placebo group. The ClinicalTrials.gov record identifies the analyses as superiority tests but does not provide an alpha-allocation or multiplicity procedure for these two comparisons. This page therefore does not assign an additional multiplicity interpretation that is not documented in the ClinicalTrials.gov record.

7. Secondary Endpoint Results: Migraine Responder Rates

Responder endpoints convert migraine outcomes into binary classifications. The registry reports 75% and 50% responder analyses for Weeks 1-12, plus a 75% responder analysis restricted to Weeks 1-4.

75% Migraine Responder Rate — Week 1-12

ComparisonMethodEffect measureEstimate95% CIP-value
300 mg ALD403 vs PlaceboCochran-Mantel-HaenszelMean Difference (Final Values)18.112.0 to 24.3<0.0001
100 mg ALD403 vs PlaceboCochran-Mantel-HaenszelMean Difference (Final Values)11.75.8 to 17.50.0001

For the 300 mg comparison, the reported estimate is 18.1, with a 95% CI of 12.0 to 24.3 and P < 0.0001. For the 100 mg comparison, the reported estimate is 11.7, with a 95% CI of 5.8 to 17.5 and P = 0.0001.

The registry labels the effect measure as Mean Difference (Final Values), despite the outcome being binary and expressed in participants. This page retains that registry terminology and does not relabel the estimates as risk ratios, odds ratios, or another effect measure.

75% Migraine Responder Rate — Week 1-4

ComparisonMethodEffect measureEstimate95% CIP-value
300 mg ALD403 vs PlaceboCochran-Mantel-HaenszelMean Difference (Final Values)21.315.0 to 27.6<0.0001
100 mg ALD403 vs PlaceboCochran-Mantel-HaenszelMean Difference (Final Values)15.39.3 to 21.4<0.0001

These are separate early-time-frame analyses from the Week 1-12 responder endpoint. The registry reports estimates of 21.3 and 15.3 for the 300 mg and 100 mg comparisons, respectively, with two-sided 95% confidence intervals of 15.0 to 27.6 and 9.3 to 21.4.

50% Migraine Responder Rate — Week 1-12

ComparisonMethodEffect measureEstimate95% CIP-value
300 mg ALD403 vs PlaceboCochran-Mantel-HaenszelMean Difference (Final Values)22.114.9 to 29.2<0.0001
100 mg ALD403 vs PlaceboCochran-Mantel-HaenszelMean Difference (Final Values)18.211.1 to 25.4<0.0001

The reported estimates are 22.1 for 300 mg ALD403 versus placebo and 18.2 for 100 mg ALD403 versus placebo. The corresponding 95% confidence intervals are 14.9 to 29.2 and 11.1 to 25.4, respectively, with P < 0.0001 for both comparisons.

How to interpret responder analyses

A responder endpoint asks a different question from a continuous change-from-baseline endpoint. Instead of estimating an average change in monthly migraine days, it classifies participants according to whether they reached the prespecified responder threshold.

The responder analyses therefore provide a categorical perspective on treatment response. However, the ClinicalTrials.gov record does not provide the underlying responder counts or percentages by arm, so those quantities should not be reconstructed from the reported mean-difference estimates.

The confidence intervals quantify uncertainty around the reported between-group estimates. The p-values address the corresponding superiority hypothesis and do not indicate the magnitude or clinical importance of the observed difference.

8. Secondary Endpoint Results: Acute Medication Days

Change in monthly acute medication days was a continuous secondary endpoint evaluated over Weeks 1-12 using ANCOVA.

ComparisonEstimate95% CIP-valueMethod
300 mg ALD403 vs Placebo-1.38-1.88 to -0.87<0.0001ANCOVA
100 mg ALD403 vs Placebo-1.15-1.66 to -0.65<0.0001ANCOVA

The outcome unit is Acute Medication Migraine Days. Both reported estimates are negative, with the 300 mg comparison estimated at -1.38 and the 100 mg comparison at -1.15.

Clinical Biostats interpretation

Because this is a change-from-baseline continuous outcome, a negative mean difference represents a more negative modeled change in monthly acute medication days in the ALD403 group relative to placebo.

The 95% CIs, -1.88 to -0.87 and -1.66 to -0.65, quantify uncertainty around those estimates. The P < 0.0001 values indicate strong evidence against the relevant null hypothesis under the reported superiority analyses, but they do not measure effect size.

9. Secondary Endpoint Results: HIT-6 Score

The Headache Impact Test (HIT-6) score was analyzed as a continuous secondary outcome from baseline to Week 12 using ANCOVA.

ComparisonEstimate95% CIP-valueMethod
300 mg ALD403 vs Placebo-2.88-3.91 to -1.84<0.0001ANCOVA
100 mg ALD403 vs Placebo-1.73-2.76 to -0.700.0010ANCOVA

The reported estimate is -2.88 for 300 mg ALD403 versus placebo and -1.73 for 100 mg ALD403 versus placebo. Their respective 95% confidence intervals are -3.91 to -1.84 and -2.76 to -0.70.

Clinical Biostats interpretation

The estimates describe between-group differences in modeled change from baseline in HIT-6 score. The negative direction indicates a more negative change in the ALD403 group than in placebo.

The p-values, <0.0001 and 0.0010, provide evidence against the corresponding null hypotheses under the reported superiority analyses. They should not be read as probabilities that the observed treatment effects are clinically meaningful.

The clinical meaning of a numerical score change requires a prespecified interpretation of the scale and a clinically important difference threshold. No such threshold is provided in the ClinicalTrials.gov record, so this page does not impose one.

10. Secondary Endpoint Results: Migraine Prevalence

Change in migraine prevalence from baseline to Week 4 was analyzed using a repeated-measures model, normalized in the ClinicalTrials.gov record as an MMRM.

ComparisonEstimate95% CIP-valueMethod
300 mg ALD403 vs Placebo-11.00-14.22 to -7.77<0.0001MMRM
100 mg ALD403 vs Placebo-8.26-11.48 to -5.05<0.0001MMRM

The outcome unit is listed as percentage of participants with migraine. The reported estimates are -11.00 for 300 mg ALD403 and -8.26 for 100 mg ALD403 versus placebo.

Clinical Biostats interpretation

The negative estimates indicate a more negative modeled change in migraine prevalence in the ALD403 groups relative to placebo. Because the outcome is expressed as a percentage of participants, the numerical difference is naturally interpreted in the scale of the reported outcome, but the registry's effect-measure field still labels it as Mean Difference (Final Values).

The MMRM framework is appropriate conceptually for repeated observations because it models the longitudinal structure rather than treating each observation as unrelated. The ClinicalTrials.gov record does not specify the covariance structure, fixed-effect terms, or missing-data assumptions, so those details are not inferred here.

The 95% confidence intervals provide precision information around the reported estimates. The P < 0.0001 values indicate evidence against the relevant superiority null hypotheses, not the magnitude or clinical importance of the effects.

11. Secondary Endpoint Results: Migraine on the Day After Dosing

The registry also reports the percentage of participants with a migraine on the day after dosing, assessed on Day 1. These binary outcomes were analyzed with the Cochran-Mantel-Haenszel method.

ComparisonTime frameMethodP-valueEstimate / CI
300 mg ALD403 vs PlaceboDay 1Cochran-Mantel-Haenszel<0.0001Not reported in registry-reported statistical analysis
100 mg ALD403 vs PlaceboDay 1Cochran-Mantel-Haenszel<0.0001Not reported in registry-reported statistical analysis
Interpretation boundary: The registry analysis provides P-values for this endpoint but no estimate or confidence interval. The result can therefore be reported as a formal posted statistical comparison with P < 0.0001 for each ALD403-versus-placebo comparison, but no numerical effect size should be inferred.

12. Results Summary

The posted statistical analyses cover one primary continuous endpoint and multiple secondary efficacy outcomes. The two primary comparisons both use ANCOVA and report negative mean differences with two-sided 95% confidence intervals that remain below zero.

Endpoint300 mg ALD403 vs Placebo100 mg ALD403 vs Placebo
Primary: Change From Baseline in Monthly Migraine Days, Week 1-12-2.60 (95% CI -3.45 to -1.74), P < 0.0001-2.03 (95% CI -2.88 to -1.18), P < 0.0001
75% Migraine Responder Rate, Week 1-1218.1 (95% CI 12.0 to 24.3), P < 0.000111.7 (95% CI 5.8 to 17.5), P = 0.0001
75% Migraine Responder Rate, Week 1-421.3 (95% CI 15.0 to 27.6), P < 0.000115.3 (95% CI 9.3 to 21.4), P < 0.0001
50% Migraine Responder Rate, Week 1-1222.1 (95% CI 14.9 to 29.2), P < 0.000118.2 (95% CI 11.1 to 25.4), P < 0.0001
Change in Monthly Acute Medication Days, Week 1-12-1.38 (95% CI -1.88 to -0.87), P < 0.0001-1.15 (95% CI -1.66 to -0.65), P < 0.0001
Change From Baseline in HIT-6 Score, Week 12-2.88 (95% CI -3.91 to -1.84), P < 0.0001-1.73 (95% CI -2.76 to -0.70), P = 0.0010
Change in Migraine Prevalence, Baseline to Week 4-11.00 (95% CI -14.22 to -7.77), P < 0.0001-8.26 (95% CI -11.48 to -5.05), P < 0.0001
Migraine on Day After Dosing, Day 1P < 0.0001; estimate not reportedP < 0.0001; estimate not reported

13. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm. It does not provide additional safety summaries in the statistical-analysis dataset, so the safety section is restricted to these reported serious adverse event counts.

Treatment armParticipants with serious adverse eventsAt risk
300 mg ALD4034350
100 mg ALD4033356
Placebo3366

300 mg ALD403

Serious adverse events affected 4 of 350 participants in the ClinicalTrials.gov record.

100 mg ALD403

Serious adverse events affected 3 of 356 participants.

Placebo

Serious adverse events affected 3 of 366 participants.

Statistical caution

These counts describe serious adverse events by arm; they do not by themselves establish causality or provide a formal between-group safety hypothesis test.

14. Statistical Methods Explained

Why was ANCOVA used for monthly migraine days?

Monthly migraine days is a continuous outcome, and the registered primary endpoint is defined as change from baseline averaged across Weeks 1-12. ANCOVA is a natural linear-model framework for comparing such outcomes while accounting for baseline information when specified in the model. The registry specifically reports ANCOVA for both primary comparisons.

What does a mean difference of -2.60 mean?

It means the reported adjusted difference in the change-from-baseline outcome between 300 mg ALD403 and placebo was -2.60. The negative direction indicates a more negative modeled change for ALD403 relative to placebo. It is not a statement that every participant experienced exactly a 2.60-day change.

Why use the Cochran-Mantel-Haenszel test for responder rates?

Responder status is binary rather than continuous. The Cochran-Mantel-Haenszel framework is designed for categorical comparisons and can account for stratification when relevant. In PROMISE-2, it was the registry-reported method for the 75% and 50% migraine responder endpoints.

Why does the registry say "Mean Difference" for a binary outcome?

The ClinicalTrials.gov record explicitly identifies the effect measure as Mean Difference (Final Values) for the responder analyses. A binary outcome would commonly be summarized with several possible effect measures, but the ClinicalTrials.gov record does not give a risk ratio or odds ratio. Retaining the registry label avoids silently replacing the reported analysis with a different one.

Why was an MMRM used for migraine prevalence?

The migraine-prevalence endpoint involves measurements over time, making repeated-measures modeling relevant. An MMRM can account for the fact that repeated observations from the same participant are correlated. The ClinicalTrials.gov record identifies the method as a Repeated Measures Model and normalizes it as MMRM, but it does not provide the full covariance or missing-data specification.

What does the p-value tell us?

A p-value measures the compatibility of the observed data with a specified null hypothesis under the statistical model. A smaller p-value provides stronger evidence against that null hypothesis. It does not quantify the size of the treatment effect, the probability that treatment is effective, or the probability that the result will be replicated.

Why report confidence intervals as well as p-values?

A confidence interval gives information about the precision of the estimated treatment difference. For example, the primary 300 mg estimate is -2.60 with a 95% CI of -3.45 to -1.74. The interval communicates considerably more information about the estimated effect than the p-value alone, although it still does not describe individual treatment responses.

15. Multiplicity and Multiple Comparisons

PROMISE-2 has one registered primary endpoint but two active-dose comparisons are reported for that endpoint: 300 mg ALD403 versus placebo and 100 mg ALD403 versus placebo. The statistical-analysis dataset also contains numerous secondary endpoints and multiple comparisons.

Analysis layerComparisons reportedStatistical issue
Primary endpoint300 mg vs placebo; 100 mg vs placeboTwo treatment comparisons for the same primary outcome
75% responder rateTwo treatment comparisonsMultiple secondary comparisons
75% responder rate, 4-week periodTwo treatment comparisonsAdditional time-frame comparisons
50% responder rateTwo treatment comparisonsAdditional secondary comparisons
Continuous secondary endpointsTwo treatment comparisons for each endpointMultiple endpoint comparisons
Day 1 migraine endpointTwo treatment comparisonsBinary secondary comparison
Multiplicity boundary: The ClinicalTrials.gov record identifies the hypothesis type as superiority and reports the individual p-values, but it does not provide an alpha-allocation strategy, hierarchical testing procedure, or multiplicity-adjustment method. Therefore, the posted p-values should be reported exactly as given rather than reinterpreted as establishing a particular familywise-error guarantee that is not documented in the ClinicalTrials.gov record.

16. Analysis Population and Causal Interpretation

The efficacy analyses use the Full Analysis Population, defined as all randomized participants who received investigational product or placebo. This is important because randomized treatment comparisons are most naturally interpreted according to the assigned treatment groups within the prespecified analysis population.

Randomization

Randomization provides the structural basis for comparing the treatment groups without intentionally assigning participants according to their expected outcomes.

Full Analysis Population

The posted efficacy analyses include randomized participants who received investigational product or placebo, according to the registry definition.

Observed vs modeled effect

The reported estimates are outputs of statistical models or categorical tests; they are not simply differences between two descriptive averages copied from the registry.

Endpoint-specific interpretation

Each estimate applies only to its specified outcome, time frame, comparison, and analysis population.

17. Blinding and Randomization

The trial is described as randomized and quadruple-masked with a parallel design. These features are central to the interpretation of the efficacy comparisons.

Why randomization matters

Randomization is intended to create treatment groups whose baseline allocation is determined by the randomization process rather than by investigator or participant choice. This supports causal interpretation of between-group efficacy differences when the trial is conducted and analyzed according to its design.

Why blinding matters

Quadruple masking is a design feature intended to reduce the influence of knowledge of treatment assignment. Blinding can be particularly relevant when outcomes involve participant-reported symptoms or decisions about clinical assessment.

Neither randomization nor masking guarantees that every source of bias is eliminated. They are design mechanisms that reduce specific sources of systematic distortion rather than statistical proof that no bias exists.

18. What the Primary Estimates Do — and Do Not — Mean

Primary 300 mg comparison

The estimate of -2.60 represents the reported mean difference for change from baseline in monthly migraine days between 300 mg ALD403 and placebo over Weeks 1-12.

It does not mean that every participant had a reduction of 2.60 migraine days, nor does it describe the full distribution of individual responses.

Primary 100 mg comparison

The estimate of -2.03 represents the corresponding reported mean difference for 100 mg ALD403 versus placebo.

It does not mean that 2.03 is the response of an individual participant, nor does the estimate alone describe how many participants benefited.

Confidence intervals

The 95% CIs of -3.45 to -1.74 and -2.88 to -1.18 quantify uncertainty around the respective estimated differences under the statistical framework. They are not prediction intervals for future individual patients.

P-values

The reported P < 0.0001 values indicate strong statistical evidence against the relevant null hypotheses under the posted analyses. They do not quantify clinical importance, treatment benefit probability, or individual response.

19. Limitations

20. Why This Trial Matters Statistically

PROMISE-2 is a useful statistical teaching case because it combines several common clinical-trial analysis strategies within a single randomized prevention study. The primary endpoint is continuous and analyzed with ANCOVA, while responder outcomes use a categorical-data method and a prevalence outcome uses a repeated-measures framework.

ConceptHow it appears in PROMISE-2
RandomizationRandomized parallel-group phase 3 design
BlindingQuadruple masking
Continuous endpoint analysisANCOVA for monthly migraine days, acute medication days, and HIT-6
Binary endpoint analysisCochran-Mantel-Haenszel analyses for responder outcomes and Day 1 migraine
Repeated measuresMMRM for change in migraine prevalence from baseline to Week 4
Confidence intervalsTwo-sided 95% CIs accompany the posted estimates where available
Hypothesis testingPosted analyses are identified as superiority tests
Multiple comparisonsTwo active-dose comparisons and numerous secondary endpoints
Analysis populationFull Analysis Population: all randomized participants who received investigational product or placebo
Safety comparisonSerious adverse events reported by treatment arm

21. Statistical Concepts in This Trial

Learn more about the methods used in this trial:

22. Related Statistical Calculators

23. Sources

Continue with the statistical methods behind PROMISE-2

Explore the underlying Clinical Biostats tutorials and statistical calculators for the methods represented in this randomized phase 3 analysis.

24. Record Summary

PROMISE-2 provides a compact example of several important principles in clinical-trial statistics. The study was a randomized, parallel-group, quadruple-masked phase 3 prevention trial with 1121 enrolled participants and three arms. Its registered primary endpoint was change from baseline in monthly migraine days, summarized in 28-day intervals and averaged across Weeks 1-12.

The two posted primary analyses compared 300 mg and 100 mg ALD403 separately with placebo using ANCOVA. The reported mean differences were -2.60 and -2.03, respectively, with two-sided 95% confidence intervals of -3.45 to -1.74 and -2.88 to -1.18, and P < 0.0001 for both comparisons.

The secondary analyses broaden the statistical picture. Responder endpoints were analyzed with the Cochran-Mantel-Haenszel method, continuous secondary outcomes with ANCOVA, and change in migraine prevalence with an MMRM. The posted results generally include estimates, two-sided 95% confidence intervals, and p-values, although the Day 1 migraine endpoint provides p-values without an estimate or confidence interval in the ClinicalTrials.gov record.

The most important statistical lesson is that these numbers answer different questions. A mean difference describes a group-level difference on a continuous scale; a responder analysis addresses a binary threshold; and an MMRM addresses repeated measurements over time. Confidence intervals describe precision, while p-values address compatibility with a null hypothesis. None of these quantities alone describes the experience of every individual participant.

Clinical Biostats methodology: A trial-results page should distinguish the reported statistical output from its interpretation. For PROMISE-2, that means preserving the registry's endpoint definitions, analysis populations, effect-measure labels, estimates, confidence intervals, and p-values while avoiding unsupported reconstruction of arm-specific results, model specifications, multiplicity procedures, or missing-data assumptions.