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
EMERGE was a randomized, parallel-group, quadruple-masked phase 3 study of aducanumab (BIIB037) in early Alzheimer's disease. The registry reports 1,643 participants, two study arms, a treatment purpose, and a primary endpoint based on change from baseline in the Clinical Dementia Rating Scale–Sum of Boxes (CDR-SB) at Week 78.
| Feature | EMERGE |
|---|---|
| Trial name | EMERGE |
| ClinicalTrials.gov identifier | NCT02484547 |
| Phase | Phase 3 |
| Condition | Alzheimer's disease |
| Study population | Early Alzheimer's disease |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 1,643 |
| Lead sponsor | Biogen |
| Sponsor type | Industry |
| Status | Terminated |
2. Clinical Question
The central statistical question was whether aducanumab produced a different change from baseline in CDR-SB score at Week 78 compared with placebo in participants with early Alzheimer's disease. The registry characterizes the hypothesis type as superiority.
Population
Participants enrolled in a phase 3 study of aducanumab for early Alzheimer's disease.
Intervention
Aducanumab (BIIB037).
Comparator
Placebo.
Primary question
Does aducanumab differ from placebo in change from baseline in CDR-SB at Week 78?
3. Trial Design
Study start
The registry lists September 15, 2015 as the study start date.
Primary completion
The registry lists August 5, 2019 as the primary completion date.
Premature study halt
The registry states that the study was halted following a prespecified futility analysis rather than for safety concerns.
4. Endpoints
Primary Endpoint
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Change From Baseline in Clinical Dementia Rating Scale - Sum of Boxes (CDR-SB) Score at Week 78 | Baseline, Week 78 | CDR-SB integrates assessments from 3 domains of cognition and 3 domains of function. Scores for the 6 domains are summed to produce a value ranging from 0 to 18, with higher scores indicating greater disease severity. A positive change from baseline indicates clinical decline. |
The registry describes CDR-SB as incorporating memory, orientation, judgment/problem-solving, community affairs, home/hobbies, and personal care. Severity anchors range from none = 0, questionable = 0.5, mild = 1, moderate = 2, and severe = 3, with the personal care domain omitting the 0.5 score. The summed score can change in increments of 0.5 or greater.
Secondary Endpoints With Posted Analyses
| Endpoint | Time frame | Outcome type |
|---|---|---|
| Change From Baseline in Mini Mental State Examination (MMSE) Score at Week 78 | Baseline, Week 78 | Continuous score |
| Change From Baseline in Alzheimer's Disease Assessment Scale-Cognitive Subscale (13 Items) (ADAS-Cog 13) at Week 78 | Baseline, Week 78 | Continuous score |
| Change From Baseline in Alzheimer's Disease Cooperative Study-Activities of Daily Living Inventory (Mild Cognitive Impairment Version) (ADCS-ADL-MCI) Score at Week 78 | Baseline, Week 78 | Continuous score |
5. Statistical Methodology
Intention-to-treat analysis
The registry defines the ITT population as all randomized participants who received at least one dose of study treatment, either aducanumab or placebo. The number of participants analyzed refers to the number analyzed for the particular outcome measure.
This definition is important because the efficacy comparisons remain anchored to the randomized treatment groups. Randomization establishes the treatment comparison before outcome information is observed, while the requirement for at least one dose defines the registry's stated ITT analysis population.
Mixed model for repeated measures
The primary CDR-SB analyses used a mixed model for repeated measures (MMRM). The dependent variable was change from baseline in CDR-SB. The model incorporated treatment group, categorical visit, treatment-by-visit interaction, baseline CDR-SB, and baseline CDR-SB-by-visit interaction. Additional covariates included baseline MMSE, Alzheimer's disease symptomatic medication use at baseline, region, and laboratory ApoE status in the low-dose analysis. The high-dose analysis also included treatment group, categorical visit, treatment-by-visit interaction, baseline CDR-SB, and baseline CDR-SB-by-visit interaction.
The same general longitudinal framework was used for the MMSE, ADAS-Cog 13, and ADCS-ADL-MCI analyses, with the corresponding baseline outcome and baseline-by-visit interaction included in the model.
The important feature is that the model evaluates the trajectory of change over repeated visits rather than reducing the entire longitudinal record to a single unadjusted comparison.
Effect measure
The registry reports the treatment effect as a difference. For the primary CDR-SB endpoint, the reported estimate is the difference between the adjusted mean change from baseline for an aducanumab group and the placebo group at the specified time point.
Two-sided confidence intervals
All posted statistical analyses use 95% two-sided confidence intervals. The confidence interval describes uncertainty around the estimated treatment difference under the fitted model and sampling framework.
Superiority testing
The registry identifies the hypothesis type as superiority. The statistical question is therefore whether the estimated treatment difference is distinguishable from the null value of no difference, rather than whether aducanumab satisfies a prespecified non-inferiority margin.
6. Results: Primary Endpoint
CDR-SB — Low-Dose Aducanumab vs Placebo
Difference in adjusted mean change at Week 78
95% CI: -0.569 to 0.041 · P = 0.0901
MMRM analysis of change from baseline in CDR-SB.
| Comparison | Estimate | 95% CI | P-value | Method |
|---|---|---|---|---|
| Placebo vs BIIB037 Low Dose | -0.26 | -0.569 to 0.041 | 0.0901 | MMRM |
| Placebo vs BIIB037 High Dose | -0.39 | -0.694 to -0.086 | 0.0120 | MMRM |
The low-dose estimate of -0.26 indicates that the adjusted mean change in CDR-SB was 0.26 score units lower with BIIB037 low dose than with placebo at Week 78 under the posted MMRM. Because a positive CDR-SB change indicates clinical decline, a negative between-group difference points toward less decline in the aducanumab group relative to placebo.
The estimate is not a statement that every participant experienced 0.26 fewer CDR-SB points of decline. It is a model-based difference in adjusted mean change between randomized groups at the specified time point.
The 95% confidence interval ranges from -0.569 to 0.041. Because the interval includes zero, the data are compatible with both a negative treatment difference and a small positive difference under the model. The interval therefore communicates more information about precision than the point estimate alone.
The P-value of 0.0901 is a measure of compatibility with the null hypothesis under the specified statistical model; it is not a measure of effect size, clinical importance, or the probability that the treatment works. The interpretation also depends on the prespecified analysis framework, repeated-measures model, analysis population, and the fact that the study was stopped following a futility analysis.
CDR-SB — High-Dose Aducanumab vs Placebo
Difference in adjusted mean change at Week 78
95% CI: -0.694 to -0.086 · P = 0.0120
MMRM analysis of change from baseline in CDR-SB.
The high-dose estimate of -0.39 indicates a lower adjusted mean change in CDR-SB with BIIB037 high dose than with placebo at Week 78. Because higher CDR-SB scores indicate greater disease severity and a positive change indicates clinical decline, the negative estimate corresponds to less estimated clinical decline relative to placebo.
The 95% confidence interval, -0.694 to -0.086, lies below zero. Within the model and analysis framework reported by the registry, the interval therefore supports a negative treatment difference rather than no difference.
The P-value of 0.0120 quantifies the statistical evidence against the null value under the specified model. It does not measure the magnitude of the treatment effect. The magnitude is described by the estimated difference and its confidence interval.
As with the low-dose analysis, this is an adjusted longitudinal estimate rather than an individual-level prediction. Interpretation should also account for the study's premature termination following a prespecified futility analysis and the distinction between statistical evidence and clinical importance.
7. Secondary Endpoint Results
MMSE
| Comparison | Estimate | 95% CI | P-value |
|---|---|---|---|
| Placebo vs BIIB037 Low Dose | -0.1 | -0.65 to 0.48 | 0.7578 |
| Placebo vs BIIB037 High Dose | 0.6 | 0.00 to 1.13 | 0.0493 |
Both analyses used MMRM for change from baseline in MMSE at Week 78. The low-dose estimate was -0.1, with a 95% CI from -0.65 to 0.48 and P = 0.7578. The high-dose estimate was 0.6, with a 95% CI from 0.00 to 1.13 and P = 0.0493.
These estimates illustrate why the direction of an effect must be interpreted in the context of the scale. Unlike CDR-SB, where higher scores indicate greater disease severity, the ClinicalTrials.gov record does not provide a directional interpretation for the MMSE score itself. The safest statistical reading is therefore to focus on the reported adjusted mean differences, their confidence intervals, and their p-values rather than assigning a clinical direction beyond what the registry explicitly states.
ADAS-Cog 13
| Comparison | Estimate | 95% CI | P-value |
|---|---|---|---|
| Placebo vs BIIB037 Low Dose | -0.701 | -1.7649 to 0.3627 | 0.1962 |
| Placebo vs BIIB037 High Dose | -1.400 | -2.4596 to -0.3396 | 0.0097 |
The low-dose comparison produced an estimated difference of -0.701, with a 95% CI of -1.7649 to 0.3627 and P = 0.1962. The high-dose comparison produced an estimated difference of -1.400, with a 95% CI of -2.4596 to -0.3396 and P = 0.0097.
The important statistical distinction is the location and width of the confidence intervals. The low-dose interval crosses zero, whereas the high-dose interval is entirely below zero. That difference describes the uncertainty in the two estimates; it should not be converted into a claim that one dose necessarily has a quantitatively larger clinical effect without a formal comparison of dose effects.
ADCS-ADL-MCI
| Comparison | Estimate | 95% CI | P-value |
|---|---|---|---|
| Placebo vs BIIB037 Low Dose | 0.7 | -0.27 to 1.73 | 0.1515 |
| Placebo vs BIIB037 High Dose | 1.7 | 0.75 to 2.74 | 0.0006 |
The low-dose comparison produced an estimated difference of 0.7, with a 95% CI of -0.27 to 1.73 and P = 0.1515. The high-dose comparison produced an estimated difference of 1.7, with a 95% CI of 0.75 to 2.74 and P = 0.0006.
8. Results Summary
| Endpoint | Low Dose vs Placebo | High Dose vs Placebo |
|---|---|---|
| CDR-SB change from baseline at Week 78 | -0.26 (95% CI -0.569 to 0.041); P = 0.0901 | -0.39 (95% CI -0.694 to -0.086); P = 0.0120 |
| MMSE change from baseline at Week 78 | -0.1 (95% CI -0.65 to 0.48); P = 0.7578 | 0.6 (95% CI 0.00 to 1.13); P = 0.0493 |
| ADAS-Cog 13 change from baseline at Week 78 | -0.701 (95% CI -1.7649 to 0.3627); P = 0.1962 | -1.400 (95% CI -2.4596 to -0.3396); P = 0.0097 |
| ADCS-ADL-MCI change from baseline at Week 78 | 0.7 (95% CI -0.27 to 1.73); P = 0.1515 | 1.7 (95% CI 0.75 to 2.74); P = 0.0006 |
The posted results show that the estimated differences are not uniform across dose levels or outcome measures. For the primary CDR-SB endpoint, the low-dose estimate was -0.26 and the high-dose estimate was -0.39. The corresponding secondary outcomes also vary in magnitude and direction across scales. This is precisely why each endpoint should be interpreted on its own statistical scale rather than combining all p-values into a single summary measure.
9. Statistical Methods Explained
Why was an MMRM used?
The outcomes in EMERGE were measured repeatedly over time, while the primary endpoint was defined as change from baseline at Week 78. An MMRM is designed for this longitudinal structure because it models repeated outcome information rather than treating observations at different visits as unrelated single measurements.
The registry's model included categorical visit and a treatment-by-visit interaction. This allows the estimated treatment difference to vary across visits rather than imposing a single constant treatment difference across the entire follow-up period.
What does a CDR-SB difference of -0.39 mean?
For the high-dose comparison, the MMRM estimated a difference of -0.39 in adjusted mean change from baseline at Week 78 relative to placebo. Because the registry states that a positive CDR-SB change indicates clinical decline, the negative value corresponds to less estimated decline in the high-dose group relative to placebo.
It does not mean that every individual participant had 0.39 fewer points of decline. It is a group-level adjusted mean difference generated by the longitudinal model.
Why is the confidence interval important?
A point estimate is only one description of the treatment comparison. The 95% confidence interval indicates the statistical precision of that estimate under the specified model. For the high-dose CDR-SB analysis, the interval from -0.694 to -0.086 is entirely below zero. For the low-dose analysis, the interval from -0.569 to 0.041 crosses zero.
The width of the interval also matters. Wider intervals indicate greater uncertainty around the estimated effect, while narrower intervals indicate greater precision.
What does the p-value tell us?
A p-value evaluates how compatible the observed data are with a specified null hypothesis under the statistical model. It does not quantify the size of the effect, the clinical importance of the effect, or the probability that the treatment hypothesis is true.
For example, the high-dose CDR-SB result has P = 0.0120, while the estimated difference is -0.39. The p-value and the effect estimate answer different questions and should be reported together rather than substituted for one another.
Why include baseline-by-visit interactions?
The registry describes baseline outcome-by-visit interaction terms in the MMRM. This allows the association between a participant's baseline score and the modeled outcome to vary across visits. In longitudinal clinical trials, this can provide a more flexible adjustment for baseline differences in the outcome trajectory than assuming a single baseline relationship at every visit.
What does ITT mean in this trial?
The registry defines ITT as all randomized participants who received at least one dose of aducanumab or placebo. The principle preserves the randomized treatment comparison within the registry's stated efficacy analysis population. It does not mean that every participant necessarily contributed the same number of repeated observations to the MMRM.
What does the premature termination mean statistically?
The registry states that the study was halted prematurely after a prespecified futility analysis and that the decision was not based on safety concerns. Early stopping can affect the amount of information accumulated for the planned analysis and therefore belongs in the interpretation of the estimated effects and their uncertainty. It also means that the observed dataset was generated under a study that did not continue according to its originally intended full course.
10. Safety
The registry reports serious adverse events by study period and arm. These counts should be kept separate from the efficacy estimates because the denominator and period can differ across safety summaries.
| Study arm / period | Serious adverse events | Affected / at risk |
|---|---|---|
| Placebo — PC Period | Serious adverse events | 81 / 547 |
| BIIB037 Low Dose — PC Period | Serious adverse events | 72 / 544 |
| BIIB037 High Dose — PC Period | Serious adverse events | 73 / 547 |
| BIIB037 Late Start: Low Dose — LTE Period | Serious adverse events | 15 / 131 |
| BIIB037 Late Start: High Dose — LTE Period | Serious adverse events | 9 / 132 |
| BIIB037 Early Start: Low Dose — LTE Period | Serious adverse events | 25 / 251 |
| BIIB037 Early Start: High Dose — LTE Period | Serious adverse events | 25 / 257 |
The post-completion-period safety groups should not be directly pooled with the primary-completion-period groups without accounting for their different periods and denominators. A count such as 25 serious adverse events among 251 participants is a descriptive safety summary for that specified group and period, not an efficacy measure.
11. Limitations
- Premature termination: the registry states that the study was halted based on a prespecified futility analysis. This changes the context in which the observed estimates should be interpreted.
- Safety was not the reason for termination: the registry specifically states that the halt was not based on safety concerns.
- Longitudinal modeling: MMRM estimates depend on the model specification, including treatment, visit, treatment-by-visit interaction, baseline outcome, and other covariates.
- Repeated measurements: participants can contribute observations at multiple visits, so the MMRM is not equivalent to a simple comparison of Week 78 means.
- Analysis population: the registry-defined ITT population requires randomized participants to have received at least one dose of study treatment.
- Multiple outcomes: the registry contains one primary endpoint and multiple secondary endpoint analyses. The individual p-values should therefore be interpreted in the context of the trial's overall endpoint structure rather than treated as unrelated tests.
- Different scales: CDR-SB, MMSE, ADAS-Cog 13, and ADCS-ADL-MCI have different scoring systems, so numerical differences should not be compared solely by their absolute magnitude.
- Safety periods: serious adverse event counts are reported for distinct study periods and denominators, limiting direct comparisons across those periods.
12. Why This Trial Matters Statistically
EMERGE provides a useful example of longitudinal clinical-trial analysis because the principal efficacy endpoint is a continuous clinical scale measured repeatedly over time rather than a binary response endpoint or a time-to-event endpoint. That changes the statistical problem substantially.
| Concept | How it appears in EMERGE |
|---|---|
| Randomization | Participants were randomized in a parallel-group phase 3 design. |
| Quadruple masking | The registry identifies the study as quadruple-masked. |
| Intention-to-treat analysis | The posted analyses define ITT as randomized participants who received at least one dose of study treatment. |
| Repeated measures | Clinical outcomes were evaluated longitudinally through an MMRM framework. |
| MMRM | The primary and posted secondary analyses used mixed models for repeated measures. |
| Continuous endpoints | CDR-SB, MMSE, ADAS-Cog 13, and ADCS-ADL-MCI were analyzed as continuous scores. |
| Confidence intervals | All posted estimates were accompanied by 95% two-sided confidence intervals. |
| P-values | Each posted comparison includes a p-value that quantifies statistical evidence under the specified model. |
| Futility analysis | The registry states that the study was halted prematurely following a prespecified futility analysis. |
| Safety by period | Serious adverse events are reported separately for primary-completion and later-term periods. |
13. Interpreting the Primary Effect Estimates
The CDR-SB high-dose estimate of -0.39 is the most direct description of the magnitude of the posted high-dose treatment difference at Week 78. It is expressed in CDR-SB score units and represents an adjusted mean difference under the MMRM.
The registry explicitly states that a positive change from baseline in CDR-SB indicates clinical decline. Consequently, a negative treatment difference indicates less modeled change toward greater disease severity relative to placebo.
The 95% CI of -0.694 to -0.086 indicates uncertainty around the high-dose estimate. The interval is relatively narrow compared with the scale of possible CDR-SB values, but it still spans a range of plausible treatment differences rather than identifying one exact effect.
The high-dose CDR-SB P-value of 0.0120 indicates statistical evidence against a null difference of zero under the posted MMRM. The p-value should not be interpreted as the probability that the null hypothesis is true or as a measure of the size of the effect.
The low-dose estimate was -0.26, with a 95% CI of -0.569 to 0.041 and P = 0.0901. The difference between this result and the high-dose result illustrates that treatment-level estimates should be examined separately rather than assuming a monotonic pattern from the numerical values alone.
14. Understanding the Secondary Results
The secondary results demonstrate an important principle in clinical biostatistics: different clinical instruments can provide different statistical signals because they measure different domains and have different scales.
MMSE
The high-dose comparison was 0.6 with a 95% CI of 0.00 to 1.13 and P = 0.0493, while the low-dose comparison was -0.1 with a 95% CI of -0.65 to 0.48 and P = 0.7578.
ADAS-Cog 13
The high-dose estimate was -1.400 with a 95% CI of -2.4596 to -0.3396 and P = 0.0097; the low-dose estimate was -0.701 with a 95% CI of -1.7649 to 0.3627 and P = 0.1962.
ADCS-ADL-MCI
The high-dose estimate was 1.7 with a 95% CI of 0.75 to 2.74 and P = 0.0006; the low-dose estimate was 0.7 with a 95% CI of -0.27 to 1.73 and P = 0.1515.
Statistical lesson
A collection of endpoint-specific estimates should not be collapsed into a single numerical "trial effect." Each outcome has its own scale, definition, uncertainty, and clinical interpretation.
15. Planned vs Observed Trial Interpretation
The registered primary endpoint was change from baseline in CDR-SB at Week 78, and the posted analyses use MMRM to estimate treatment differences at that time point. The registry also identifies the hypothesis as superiority.
At the same time, the registry states that the study was terminated after a prespecified futility analysis. That fact is important because trial stopping is part of the data-generating process. A statistical result should not be interpreted as though the trial necessarily continued unchanged to its originally intended conclusion.
16. Sources
- ClinicalTrials.gov: NCT02484547 — EMERGE.
- PubMed record: PMID 40545559.
- PubMed record: PMID 40413110.
- PubMed record: PMID 38165296.
- PubMed record: PMID 36600109.
- PubMed record: PMID 34807243.
Continue through the Clinical Biostats knowledge graph
Explore the statistical concepts represented in EMERGE, from randomized trial design and confidence intervals to longitudinal MMRM analysis and interpretation of p-values.
17. Related Tutorials
Learn more about the methods used in this trial:
18. Related Calculators
19. Record Summary
EMERGE provides a detailed example of longitudinal clinical-trial analysis in which continuous clinical outcomes were repeatedly assessed and analyzed using MMRM. The primary endpoint was change from baseline in CDR-SB at Week 78, with posted adjusted mean differences of -0.26 for BIIB037 low dose versus placebo and -0.39 for BIIB037 high dose versus placebo. The corresponding 95% confidence intervals and p-values show materially different levels of statistical evidence between the two comparisons.
The secondary MMSE, ADAS-Cog 13, and ADCS-ADL-MCI analyses further demonstrate why clinical-trial interpretation should preserve endpoint definitions and scales. Across these measures, the registry reports distinct effect estimates, confidence intervals, and p-values for the low- and high-dose comparisons.
The most important statistical context is that the study was terminated after a prespecified futility analysis and that the registry states the decision was not based on safety concerns. Consequently, the numerical results are best understood together with the randomized design, ITT definition, longitudinal MMRM framework, repeated-measures structure, and early study termination.