A Safety and Efficacy Study to Evaluate AMG 334 in Migraine Prevention: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Safety and Efficacy Study to Evaluate AMG 334 in Migraine Prevention (NCT02630459), a randomized phase 2 study. The primary outcome is Change From Baseline in Mean Monthly Migraine Days at Months 4, 5 and 6.

2. Clinical Question

3. Trial Design

This was a Phase 2, randomized, double-blind, placebo-controlled, parallel-group study of AMG 334, also called erenumab, for migraine prevention. The registry lists triple masking. The study had a double-blind treatment phase followed by an open-label treatment phase. The registry record describes the study as multicenter and lists a total enrollment of 475.

During the double-blind treatment phase, the listed groups received placebo or erenumab at 28 mg, 70 mg, or 140 mg by subcutaneous injection on day 1 and at weeks 4, 8, 12, 16, and 20. The primary endpoint used a 4-week baseline phase and measurements during months 4, 5, and 6 of the double-blind phase. The posted primary analysis includes 136 participants in the placebo group, 66 in the 28 mg group, 135 in the 70 mg group, and 136 in the 140 mg group.

The subsequent open-label treatment phase used erenumab at 70 mg and/or 140 mg, depending on visit completion status after approval of a protocol amendment. Its stated duration was 76 weeks. The registry also describes a device-use sub-study among a subset of participants in the open-label phase. These later phases are distinct from the placebo-controlled primary comparison and should not be read as randomized, blinded comparisons against placebo.

The registry lists the study as completed, with a start date of 2016-01-06 and primary completion date of 2017-09-25. It does not report the sample-size justification or further detail about the randomization procedure in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: OLTP: participants received erenumab 70 mg and/or 140 mg QM SC (depending on the participant's visit completion status after a protocol amendment). Optional 85-day clinical home use (CHU) sub-study during the OLTP: participants randomized 1:1 to erenumab using two 70 mg/mL or one 140 mg/mL autoinjector (AI)/pen(s).

Only the first 6 of 8 groups are shown.

Double-Blind Treatment Phase (DBTP)

Reasons for not completing this period:

Open-Label Treatment Phase (OLTP)

Reasons for not completing this period:

Optional CHU Sub-Study (During OLTP)

Analysis population for the primary outcome, as defined in the registry: Efficacy Analysis Set: Participants who received at least 1 dose of the investigational product (IP) and had at least 1 change from baseline measurement in monthly migraine days during the DBTP.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline in Mean Monthly Migraine Days at Months 4, 5 and 6, measured over 4-week baseline phase and months 4, 5 and 6 of DBTP..

The primary endpoint was change from baseline in mean monthly migraine days at months 4, 5, and 6 of the double-blind treatment phase. A migraine day was defined as a calendar day with a qualified migraine headache: migraine with or without aura lasting at least 30 minutes and meeting specified pain-feature or associated-symptom criteria. The baseline value was the number of migraine days during the 4-week baseline phase. Change was calculated as the mean monthly migraine days across months 4, 5, and 6 minus the baseline number.

This endpoint represents a continuous measure: it describes the average change in migraine days, rather than simply whether a participant met a response threshold. A group average can capture shifts across the full distribution of measured days, while a responder outcome summarizes whether participants crossed a chosen threshold. Both perspectives can be informative, but they answer different questions.

The registry reports least-squares means estimated with a generalized linear mixed model. This model-based approach incorporates treatment, visit, the treatment-by-visit interaction, a stratification factor, and baseline value. The endpoint is useful for comparing average change over the specified treatment-period visits while accounting for the listed model terms. It does not by itself describe the range of individual participant outcomes or establish how well the average represents people outside the analysis population.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary analysis used a generalized linear mixed model to estimate least-squares means and between-group differences. The registry lists treatment, visit, treatment-by-visit interaction, stratification factor, and baseline value as model terms. The stratification factor was prior/current treatment status, categorized as current, prior only, or no prior/current treatment. The registry does not report further model details, such as the link function, covariance structure, or procedures for missing observations.

What does a least-squares mean represent?

A least-squares mean is a model-based estimate of a group mean after accounting for the terms included in the model. It is not necessarily the same as the simple arithmetic average of the observed values. In this study, the estimates concern change in monthly migraine days over the specified visits, with the listed baseline and treatment-related terms included. The corresponding between-group estimate compares these modeled changes.

How should the confidence intervals be read?

A confidence interval expresses uncertainty around an estimate under the statistical model and analysis assumptions. A narrower interval indicates greater precision than a wider interval, all else equal. The 95% intervals reported for the primary comparisons lie below zero. This is consistent with the registry’s reported superiority classifications and p-values. A confidence interval does not mean that the true effect has a 95% probability of falling within that particular interval; it is a property of the estimation procedure across repeated use under its assumptions.

The primary model accounts for repeated visits through its treatment, visit, and treatment-by-visit terms, as well as the other listed covariates. Repeated measurements from one participant are generally related rather than independent. Mixed models can represent such structured data, but their behavior depends on model specification and assumptions. The registry record does not give enough detail to assess the full specification or evaluate those assumptions independently.

What do the secondary binary comparisons estimate?

One posted secondary outcome is the percentage of participants with at least a 50% reduction from baseline in mean monthly migraine days at months 4, 5, and 6. This converts a continuous measure into a responder classification. The posted simple recalculations compare the responder proportions with risk differences, confidence intervals, and unadjusted p-values for the listed pairings. A risk difference expresses the difference between two proportions on the probability scale; it is not the same estimand as an odds ratio.

The sponsor’s reported analyses for this responder outcome use a Cochran-Mantel-Haenszel method and report common odds ratios for the placebo comparison with the 70 mg and 140 mg groups. An odds ratio compares odds rather than proportions, and the method can account for stratification. The registry record does not specify all implementation details for these analyses. Since the recalculated risk differences and sponsor-reported common odds ratios use different effect measures, and not all pairings are the same, their numerical estimates should not be treated as direct duplicates.

What do the planning illustrations mean?

The sample-size illustrations are based on observed effects, not on the trial’s own design justification. The stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation, with the effect to detect set at the stated fraction of the observed effect. For the 70 mg versus placebo responder comparison, the illustration gives 50 participants per arm for 80% power and 66 participants per arm for 90% power when planning for 100% of the observed effect. For 75% of that effect, the corresponding figures are 78 and 104; for 50%, they are 150 and 200. These are planning illustrations only, not observed power and not evidence that the trial was designed using these assumptions.

In general, smaller target effects require larger samples to distinguish them from random variation with the same design assumptions. Such illustrations are sensitive to the assumed control response, treatment effect, allocation, analysis, and significance level. They do not account for every practical feature of a real trial unless those features are explicitly included. The registry record does not report whether these assumptions match the sponsor’s original sample-size planning.

Why can over-dispersion matter?

As a general statistical point, over-dispersion occurs when variability in observed data exceeds what a model’s assumed distribution allows. If present and not handled appropriately, standard errors and confidence intervals may be too optimistic or otherwise poorly calibrated. The primary endpoint is a count-like measure summarized as monthly migraine days, but the registry record does not report a dispersion assessment or enough model details to determine how this issue was evaluated. It is therefore a general consideration, not a claim that over-dispersion occurred in this trial.

The reported p-values quantify how incompatible the observed data are with a specified null hypothesis under the analysis assumptions. They do not measure the size or clinical importance of an effect, and they do not give the probability that the null hypothesis is true. The estimates and their confidence intervals are needed to understand direction, magnitude, and precision alongside the p-values.

10. Analyses Reported by the Sponsor

The primary endpoint comparisons reported by the sponsor use a generalized linear mixed model and are the principal inferential results for the placebo-controlled primary question. The posted primary outcomes include group-specific least-squares means and sponsor-reported model-based differences, confidence intervals, and p-values. The record does not include a separate simple recalculation for these continuous comparisons, so there is no unadjusted primary recalculation to treat as a competing estimate.

For the posted responder outcome, the sponsor reports common odds ratios using a Cochran-Mantel-Haenszel analysis for the 70 mg and 140 mg placebo comparisons. Simple recalculations use risk differences and unadjusted p-values, and include a 28 mg comparison not listed among the sponsor-reported analyses for that outcome. Because odds ratios and risk differences are different measures, and because the comparison sets differ, the recalculations provide a supplementary unadjusted view rather than a reproduction of the sponsor’s analysis.

The unadjusted calculations use posted summary counts and do not incorporate the sponsor’s model structure or stratification in the same way. The sponsor-reported methods should therefore be used to understand the trial’s stated inferential comparisons; the recalculations are useful for orientation, not substitutes for those analyses.

11. Multiplicity, Missing Data and Analysis Populations

The registry identifies the primary endpoint and reports several primary comparisons against placebo, along with a range of secondary outcomes. It does not report a statistical plan describing how type I error was controlled across those comparisons or across the secondary outcomes. Without that information, it is not possible to determine from the registry record whether a formal multiplicity procedure was prespecified or how it was applied.

As a general statistical issue, evaluating multiple hypotheses increases the chance that at least one result will appear statistically notable by chance if each test is interpreted separately. Multiplicity procedures can manage that risk, but they can also affect the interpretation and precision of individual tests. The reported p-values should therefore be read as the values associated with the listed analyses, not as proof that all possible comparisons were accounted for in a single error-control strategy.

The primary analysis population is defined as participants who received at least one dose of investigational product and had at least one change-from-baseline measurement in monthly migraine days during the double-blind treatment phase. This is the registry’s stated efficacy analysis set. The record does not report the number of randomized participants excluded from that set, the reasons for any exclusions, or a complete participant flow for the primary analysis.

The registry record also does not report how missing endpoint data were handled, what assumptions were made about missingness, or whether sensitivity analyses examined alternative assumptions. These details matter because incomplete follow-up can affect model-based estimates, particularly if the availability of measurements is related to outcomes or treatment. Their absence limits independent assessment of the robustness of the primary results.

Interim analyses, stopping rules, and the detailed statistical analysis plan are not reported in the registry record. The reported model terms are available, but the record does not provide all technical specifications needed to reproduce the analysis. These omissions do not by themselves establish that the analyses were inappropriate; they define what cannot be evaluated from the registry record alone.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: mycobacterial infection (4.2%), hand fracture (1.5%), systemic lupus erythematosus (0.7%), prinzmetal angina (0.7%), haemorrhoids (0.7%).

During the double-blind treatment phase, the registry reports serious adverse events in 4 of 136 participants in the placebo group, 1 of 66 in the 28 mg group, 1 of 135 in the 70 mg group, and 1 of 137 in the 140 mg group. It reports other adverse events in 55 of 136, 28 of 66, 67 of 135, and 58 of 137 participants, respectively. No deaths are listed for these groups.

For the open-label treatment phase, serious adverse events are reported in 18 of 386 participants in the 70 mg group, 9 of 189 in the 140 mg group, and 27 of 459 in the combined 70/140 mg total. Other adverse events are reported in 231 of 386, 133 of 189, and 343 of 459 participants, respectively. No deaths are listed for these groups. The open-label phase is not a concurrent placebo-controlled comparison, and its groups should not be directly compared with the double-blind groups as though they came from the same randomized contrast.

These are participant counts, not measures of event severity, duration, causality, or comparative risk adjusted for follow-up. The registry lists selected serious-event terms, but the counts alone do not establish whether an event was caused by study treatment. The registry record does not report enough detail to evaluate individual event narratives or differences in time at risk.

13. What the Result Does, and Does Not, Mean

The primary results describe model-estimated average changes in monthly migraine days during the double-blind treatment phase for the efficacy analysis set. The sponsor-reported comparisons indicate that each listed erenumab dose group had a lower modeled change than placebo, with confidence intervals below zero and the reported p-values shown for those comparisons. The estimates provide information about average group differences and their statistical precision under the stated model.

They do not establish the exact experience of any individual participant. Group averages can coexist with a wide range of individual outcomes, and the registry does not report the full distribution of participant-level changes in the information. The estimates also do not establish how the results would apply to populations or settings not represented in the analyzed participants.

The results do not, by themselves, determine the clinical importance of the estimated differences. Statistical evidence addresses compatibility with a null comparison under model assumptions; clinical importance requires context beyond a p-value and is not defined by the registry’s reported significance tests. The record does not provide a threshold for a clinically important difference in monthly migraine days.

Nor do the primary placebo comparisons establish that one erenumab dose is better than another. Each reported primary contrast is against placebo; no direct dose-to-dose primary comparison is included. The open-label phase has a different design from the blinded comparison and cannot be treated as a placebo-controlled estimate of comparative effect.

Finally, results should be considered alongside the analysis-population definition, missing-data details that are not reported in the registry record, and the breadth of secondary outcomes. The statistical results describe what the posted analyses estimate; they are not a substitute for a full assessment of the study protocol, analysis plan, participant-level data, or other evidence.

14. Why This Trial Matters Statistically

This trial illustrates why endpoint definition matters in clinical research. Monthly migraine days were not defined as any headache day; the registry specifies criteria for a qualified migraine headache, including duration and pain or associated-symptom features. That operational definition determines what is counted and therefore shapes the outcome being analyzed.

It also distinguishes a continuous endpoint from a responder endpoint. The primary analysis estimates mean change, while the 50% reduction outcome classifies participants according to whether they cross a threshold. These summaries can support different interpretations and should not be treated as interchangeable. Presenting both can help readers see how an average difference relates to a threshold-based response, while still recognizing that a threshold discards some information about the amount of change.

The study further demonstrates the distinction between a model-based sponsor analysis and a simple unadjusted recalculation. Model terms and stratification can affect estimates and uncertainty; recalculations from summary counts answer a more limited question. Keeping the estimand and method clear helps prevent accidental comparisons between risk differences and odds ratios or between analyses using different group pairings.

Finally, the registry record shows why statistical interpretation depends on more than a favorable p-value. Confidence intervals convey precision, the analysis population defines whose data contribute, and missing-data and multiplicity plans affect how results should be read. When those details are not reported, readers can interpret the posted estimates but cannot fully assess every analytic assumption or the robustness of the overall evidence.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

This is a Phase 2 study, and the registry record does not provide enough participant-characteristic detail to assess how representative the analyzed population was. The primary efficacy analysis set required both exposure to investigational product and at least one post-baseline measurement. That definition means the reported estimate is not automatically an estimate for every randomized participant; the number and reasons for any exclusions from the efficacy set are not reported in the registry record.

The registry does not report the handling of missing primary endpoint measurements, the assumptions underlying that handling, or sensitivity analyses. It also does not provide a complete technical specification of the generalized linear mixed model. These gaps limit independent evaluation of whether the estimates are robust to alternative modeling or missing-data assumptions.

Multiple primary and secondary comparisons are reported, but the registry record does not report a multiplicity strategy. This makes it difficult to evaluate the family-wise interpretation of the reported p-values across outcomes and comparisons. The secondary outcome record is also incomplete in the summarized registry results: detailed numerical findings for most listed secondary domains are not reported in the registry record.

The open-label phase followed the blinded phase and used erenumab rather than a concurrent placebo comparison. Its safety counts describe participants in that phase, but they do not provide the same randomized contrast as the double-blind treatment phase. Safety counts also do not by themselves characterize event timing, duration, severity, or causal attribution. The registry record does not report a sample-size justification, interim-analysis plan, or the full statistical analysis plan, so those aspects cannot be assessed from the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: A Safety and Efficacy Study to Evaluate AMG 334 in Migraine Prevention — NCT02630459.
• Registry record last updated 2022-10-12; status last verified 2022-10.
• Results first posted 2019-03-13.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

The registry record describes a completed Phase 2 randomized, triple-masked, parallel-group study of erenumab for migraine prevention, with placebo control during the double-blind treatment phase and a later open-label phase. Total enrollment was 475. The primary endpoint was change from baseline in mean monthly migraine days at months 4, 5, and 6.

The sponsor reported model-based comparisons of each erenumab dose group with placebo. All three reported primary contrasts had 95% confidence intervals below zero, with p = 0.004 for 28 mg and p < 0.001 for 70 mg and 140 mg. These estimates apply to the stated efficacy analysis set and depend on the reported model.

The record also lists secondary efficacy and safety outcomes, but detailed results for most secondary outcomes and several analysis-plan details are not reported in the registry record. The posted estimates are best read with attention to the endpoint definition, analysis population, comparison method, and uncertainty.