An Efficacy and Safety Trial of Verubecestat (MK-8931) in Mild to Moderate Alzheimer's Disease (P07738): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for An Efficacy and Safety Trial of Verubecestat (MK-8931) in Mild to Moderate Alzheimer's Disease (P07738) (NCT01739348), a randomized phase 2, phase 3 study. The primary outcome is [Part I (Base Study)] Change From Baseline in the Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) Score.

2. Clinical Question

The registry lists Alzheimer's disease as the condition and describes the study as randomized, double-blind, and parallel-group. The trial included 2,211 participants overall, but that enrollment count should not be confused with the endpoint-specific analysis counts. The primary outcome was assessed at baseline and week 78. The sponsor's comparison was between each of the two reported verubecestat dose arms and the placebo arm. Arm C has no estimable primary endpoint value in the posted results, and it was excluded from the specified primary analysis population.

3. Trial Design

This randomized, placebo-controlled trial was described as Phase 2 and Phase 3, double-blind, and parallel-group. It enrolled 2,211 participants with mild to moderate Alzheimer's disease. Part I was designed to assess efficacy and safety over 78 weeks. Part II was a long-term double-blind extension for participants who completed Part I and chose to continue, with treatment for up to an additional 260 weeks.

During Part I, Arm A received verubecestat 12 mg once daily, Arm B received 40 mg once daily, and Arm D received placebo once daily. Arm C received 60 mg until the first IA, then 40 mg for the remainder of Part I. In Part II, the listed continuing regimens were verubecestat 12 mg for Arm A and 40 mg for Arms B, C, and D. The placebo comparison therefore applies to Part I; Part II was not a continued placebo-controlled comparison for Arm D.

The study summary identifies two primary hypotheses for Part I: superiority to placebo on change from baseline in ADAS-Cog and on change from baseline in ADCS-ADL, each at week 78. The results discussed for the primary endpoint concern ADAS-Cog. The registry record reports a termination status, but the reason for termination is not reported in the registry record. The protocol's detailed allocation procedures, sample-size justification, and any planned interim-analysis procedures are not reported in the registry record. The record identifies a Safety Cohort consisting of the first approximately 400 participants entering Part I, but does not provide further design details about that cohort in the information.

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: 2211 participants were randomized in Part I, with 2210 receiving treatment. As planned per protocol, no participants were randomized to the Verubecestat 60 mg arm (Arm C) in the Main Cohort. Participants completing Part I were eligible to continue to Part II. The trial was terminated early and did not complete as planned.

Part I (Base Study)

Reasons for not completing this period:

Part II (Extension Study)

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized participants with a baseline and ≥1 within-analysis-window ADAS-Cog observation subsequent to ≥1 dose of study drug (FAS population). Per protocol, the 200 participants enrolled before IA (all arms) and all participants receiving 60mg Verubecestat (Arm C) were excluded.

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.

Population: All randomized participants in Study Part I (Base Study)

6. Primary Endpoint

The registry lists the primary outcome as [Part I (Base Study)] Change From Baseline in the Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) Score, measured over Baseline and week 78.

The primary endpoint was change from baseline in the Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog), assessed at baseline and week 78. The scale combines 11 cognitive measures, including speech, comprehension, word finding and recall, naming, orientation, and memory-related tasks. Individual item scores contribute to a total score ranging from 0-70. Higher total scores indicate greater cognitive impairment and Alzheimer's disease severity; an increase over time reflects worsening on this scale.

The registry reports least squares mean change from baseline. A least squares mean is a model-based estimate, rather than simply the arithmetic average of observed scores. This approach can be useful for comparing groups while accounting for factors included in the statistical model. The particular model terms and covariates beyond the named longitudinal ANCOVA method are not reported in the registry record.

ADAS-Cog provides a standardized way to quantify cognitive performance for a trial endpoint. Its direction matters when interpreting the estimates: positive change indicates higher scores, and higher scores indicate more impairment. The posted arm estimates are positive in all three arms with a reportable value. The endpoint's score direction describes the measure; it does not by itself establish why scores changed or what an individual participant experienced.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

General explanation: A continuous endpoint is measured on a numeric scale, such as ADAS-Cog. A treatment comparison can be expressed as a difference between estimated group means. The sign of that difference depends on the order of the groups in the comparison, while the scale's direction determines what higher or lower scores mean clinically. For ADAS-Cog, higher scores indicate greater impairment.

What does longitudinal ANCOVA contribute?

The sponsor reported a longitudinal ANCOVA for the primary comparisons. In general, analysis of covariance uses a statistical model to estimate group differences while accounting for measured variables included in the model. A longitudinal approach is designed for repeated measurements over time. The posted primary summary identifies the method but does not report the model's full specification, covariates, covariance structure, or detailed handling of incomplete repeated measurements. Those details are not reported in the registry record.

The reported least squares means are model-based estimates of change from baseline. They are not necessarily identical to raw averages calculated directly from participant-level changes. The registry reports the estimated contrast between each reportable verubecestat arm and placebo, along with confidence intervals and p-values. For each contrast, the confidence interval describes a range of values compatible with the model and observed data under the method's assumptions. It is not a range containing a known fixed proportion of individual participant outcomes.

How should the confidence intervals and p-values be read?

The arm-level estimates are accompanied by 95% confidence intervals. The sponsor's reported pairwise comparisons use 97.51% confidence intervals. The registry record does not explain why that confidence level was selected, so its rationale should not be inferred. A confidence interval that includes zero indicates that a zero difference is among the values compatible with the reported interval. It also leaves room for effects in either direction; it does not show that the true difference is exactly zero.

A p-value measures how unusual the observed data, or data more extreme, would be under a specified null model and its assumptions. It is not the probability that the null hypothesis is true, nor a measure of the size or importance of an effect. The p-values reported for the primary comparisons should be read alongside the estimated differences and their intervals. The registry's multiple primary hypotheses and outcome measures also make the multiplicity plan relevant; that plan is not reported in the registry record.

Why does the analysis population matter?

The primary analysis population was not simply everyone enrolled. It included randomized participants with a baseline and at least one within-analysis-window ADAS-Cog observation after at least one dose. The protocol description also excludes the 200 participants enrolled before IA and all participants receiving 60 mg verubecestat in Arm C. Therefore, the primary comparison applies to the stated analysis population, not automatically to every randomized participant or everyone enrolled. The registry gives no further detail about reasons for missing endpoint measurements or the assumptions used for missing data.

What do the secondary binary calculations represent?

The posted responder outcome is binary: each participant either met the responder definition or did not. A risk difference expresses the difference between the two groups' event proportions. Its confidence interval reflects uncertainty in that estimated difference. The recalculated unadjusted comparisons in the secondary results use summary counts and are distinct from the sponsor's primary longitudinal ANCOVA. They should not be treated as adjusted sponsor analyses. The registry record lists no sponsor-reported analysis for this secondary outcome.

General explanation: Over-dispersion occurs when binary outcomes vary more than a simple statistical model expects. If present, treating observations as less variable than they are can make uncertainty appear too small. The posted planning illustration describes a normal approximation and does not state that it adjusts for over-dispersion; no conclusion about over-dispersion in this trial can be drawn from the registry record.

How should the planning illustrations be used?

The sample-size illustrations are planning exercises based on fractions of the observed secondary effect, with two-sided significance level 0.05, equal allocation between two arms, and a normal approximation. They are not the trial's own design justification and are not observed power. In general, a planning calculation estimates the participants per arm that a hypothetical future study would require under specified assumptions. It does not change the uncertainty of the completed trial or show that the trial had a particular probability of detecting an effect. The registry's own sample-size justification is not reported in the registry record.

10. Analyses Reported by the Sponsor

The registry's primary comparisons are the sponsor-reported longitudinal ANCOVA results. For Arm A versus placebo, the estimated difference in least squares means was 0.2, with a 97.51% confidence interval from -0.9 to 1.3 and p = 0.629. For Arm B versus placebo, it was 0.4, with a 97.51% confidence interval from -0.8 to 1.5 and p = 0.463. These are adjusted model-based comparisons, not simple subtractions of the displayed arm estimates.

No discrepancy is flagged between the sponsor's analysis and the recalculated results. The sponsor's specified analysis is the relevant primary result because it reflects the reported longitudinal ANCOVA and analysis population. A simple recalculation from summary statistics, where available for secondary outcomes, is not a replacement for that model. The posted information does not give participant-level data or enough model details to reproduce the primary ANCOVA independently.

For the primary endpoint, the reported confidence intervals include zero and the reported p-values do not provide evidence against a zero difference under the sponsor's analysis. This should not be interpreted as proof that the treatments are equivalent. The registry does not report an equivalence or non-inferiority objective for this endpoint.

11. Multiplicity, Missing Data and Analysis Populations

The study summary identifies two primary hypotheses for Part I: superiority to placebo on ADAS-Cog change and superiority to placebo on ADCS-ADL change at week 78, with at least one verubecestat dose required to be superior for each measure. The registry also lists multiple secondary outcome measures and analyses. When several hypotheses are tested, multiplicity is a statistical concern because the chance of at least one false-positive finding can increase across the family of tests. The registry record does not report the prespecified multiplicity strategy, the order in which hypotheses were tested, or how the primary hypotheses were linked for decision-making. The reported 97.51% comparison intervals are not accompanied by an explanation of their rationale.

The endpoint analysis required a baseline measurement and at least one subsequent ADAS-Cog observation within the analysis window after at least one dose. It also excluded the 200 participants enrolled before IA and all participants receiving 60 mg verubecestat in Arm C. This makes the primary analysis population narrower than the total enrollment. The record does not report how many participants were excluded for each reason, how much outcome data were missing, or the reasons for missingness.

Missing outcomes can affect interpretation if the participants with observed measurements differ from those without them. General statistical methods can address missing data under assumptions, but the assumptions and approach used for this endpoint are not reported in the registry record. The details of the analysis window, sensitivity analyses, and any methods for handling intercurrent events are also not reported in the registry record.

The study summary identifies a Safety Cohort and an interim-analysis-related treatment change in Arm C, but it does not provide enough detail to infer the full interim analysis plan or its statistical consequences. A sample-size justification, planned power, and any alpha-spending or other adjustment are not reported in the registry record. These omissions limit assessment of how the final analyses relate to the original statistical plan; they do not by themselves establish that the analysis was inappropriate.

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: pneumonia (4.9%), anaemia (3.3%), hip fracture (3.3%), osteoarthritis (3.3%), squamous cell carcinoma (3.3%).

The safety record reports serious adverse events, other adverse events, and deaths separately for Part I and Part II treatment groups. For Part I, serious adverse events were reported among 134 of 702 participants in Arm A, 162 of 700 in Arm B, 24 of 103 in Arm C, and 121 of 705 in Arm D. The corresponding counts for other adverse events were 408, 440, 66, and 337. Deaths reported for these groups were 9, 13, 3, and 6, respectively.

Part II figures are listed separately for continuing treatment groups, with their own numbers at risk. Part II participants had completed Part I and chose to continue, so these groups should not be read as directly comparable to the original randomized Part I groups without further information about selection and follow-up. The registry lists pneumonia at 4.9% among its top serious adverse event terms; the other listed terms are anaemia, hip fracture, osteoarthritis, squamous cell carcinoma, and syncope, each at 3.3%. The record excerpt does not specify a denominator for those term percentages.

These are descriptive counts, not an adjusted comparison of risks or proof of causation. The counts alone do not account for duration of exposure, differences in follow-up, or the nature and severity of individual events. The registry information is not a complete clinical assessment of benefits and harms.

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

The primary result describes change on a cognitive scale over the Part I period in a defined analysis population. The posted least squares mean changes were 7.9 in Arm A, 8.0 in Arm B, and 7.7 in placebo Arm D. Since higher ADAS-Cog scores indicate greater impairment, the values describe increases in impairment scores from baseline on average as estimated by the model. The arm comparisons are more informative for the treatment question than comparing each arm's change with zero, because they directly compare verubecestat with placebo.

The estimated differences versus placebo were 0.2 and 0.4, and both reported confidence intervals included zero. This means the analysis does not distinguish those estimates from zero at the stated interval level. It does not establish that there is no effect, that the groups are equivalent, or that any difference is clinically unimportant. The registry record does not provide a threshold for clinical importance for this endpoint.

The result applies to the week 78 endpoint and the stated population; it is not a summary of all participants, all outcomes, or the long-term extension. It also does not resolve the ADCS-ADL co-primary question or establish the pattern across secondary outcomes whose numerical results are not reported in the registry record. Safety counts provide descriptive context but should not be treated as a complete comparative safety analysis.

Statistical interpretation is constrained by information not reported in the registry record, including the full analysis model, missing-data approach, multiplicity plan, and reason for termination. The findings should be read as the posted estimates and their uncertainty, rather than as a definitive statement about treatment efficacy or clinical use.

14. Why This Trial Matters Statistically

This trial illustrates why an endpoint estimate should not be interpreted from its point value alone. The ADAS-Cog arm estimates are all positive, but the direct treatment contrasts and their confidence intervals address whether the groups differed. In this case, the reported comparison intervals include zero, and the p-values do not provide evidence against a zero difference under the sponsor's analysis. A numerical ordering of estimates is not, by itself, evidence of a reliable treatment difference.

The design also highlights the importance of defining the analysis population. The total enrollment was 2,211, while the primary endpoint population had additional measurement and exposure requirements and explicitly excluded specified participants. Understanding who contributes to an estimate is essential when relating a result to the full randomized population.

Several layers of evidence need to remain distinct: the primary longitudinal model, descriptive endpoint estimates, secondary outcomes, planning illustrations, and adverse-event counts answer different questions. The secondary binary calculations are unadjusted, the planning calculations are hypothetical rather than observed power, and safety counts are descriptive. Treating these as interchangeable would overstate what each can support.

Finally, the record shows a long-term extension and a terminated status, but the reason for termination and detailed statistical-plan information are not reported in the registry record. This makes transparent separation of reported results from unresolved design and analysis details especially important when communicating the trial.

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.

The registry record does not report the reason for termination. It therefore cannot establish whether termination was related to efficacy, safety, operational considerations, or another cause. The primary analysis is defined for participants with an observed baseline and at least one subsequent within-window measurement after dosing, with additional exclusions specified; the extent and reasons for missing measurements are not reported in the registry record.

Although the sponsor names longitudinal ANCOVA, detailed model specifications are not reported. The record does not provide the covariates, repeated-measure structure, missing-data assumptions, or sensitivity analyses needed to assess the model in detail. The multiplicity plan for the two primary hypotheses and the wider set of outcomes is also not reported. These limitations constrain independent assessment of the analysis strategy.

The primary outcome information includes reportable values for Arms A, B, and D, but no estimable value for Arm C. The endpoint description states that Arm C was excluded from the specified primary population, so the reported primary comparisons do not characterize its endpoint outcome. Numerical results for most listed secondary endpoints are not reported in the registry record, limiting assessment of whether other outcome domains align with ADAS-Cog.

Part II was an extension open to participants completing Part I who chose to continue. Its results cannot be assumed to retain the same randomized comparison as Part I, and the registry information does not provide the details needed to evaluate selection into the extension. The serious adverse event and other event counts are descriptive; exposure duration, event-level details, and a complete adjusted safety comparison are not reported in the information considered. These are limits on interpretation, not evidence that unreported outcomes were favorable or unfavorable.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: An Efficacy and Safety Trial of Verubecestat (MK-8931) in Mild to Moderate Alzheimer's Disease (P07738) — NCT01739348.
• Registry record last updated 2018-10-24; status last verified 2018-09.
• Results first posted 2018-05-16.
• 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

NCT01739348 is a randomized, double-blind, parallel-group Phase 2 and Phase 3 trial of verubecestat in mild to moderate Alzheimer's disease. The study enrolled 2,211 participants and included a 78-week placebo-controlled Part I followed by an optional long-term double-blind extension. The registry status is terminated; the reason is not reported in the registry record.

The primary endpoint discussed is ADAS-Cog change from baseline at week 78. The sponsor-reported longitudinal ANCOVA comparisons of Arms A and B with placebo have confidence intervals that include zero. The analysis population excluded specified participants and required an observed baseline and subsequent outcome measurement. Multiplicity, missing-data methods, and detailed model specifications are not reported in the registry record. Secondary outcome coverage and adverse-event counts add context but do not replace the primary analysis or establish a complete overall conclusion.