Cognition Platform Study in Participants at Risk for Alzheimer's Disease (AD) (MK-0000-413): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Cognition Platform Study in Participants at Risk for Alzheimer's Disease (AD) (MK-0000-413) (NCT04730635), a randomized phase 1 study. The primary outcome is Change From Baseline in Averaged Correct Response Rate on the One Card Learning Task to Week 8.

2. Clinical Question

3. Trial Design

This was a randomized, triple-masked, parallel-group Phase 1 study sponsored by Merck Sharp & Dohme LLC. The registry lists 44 participants enrolled and the study status as completed. The recorded study start was 2021-03-23 and the primary completion date was 2023-01-20. The conditions listed are Alzheimer's disease and mild cognitive impairment.

Participants were assigned to donepezil or placebo. Donepezil was given orally once daily in a scheduled titration, at doses up to 10 mg, for 56 days. Placebo was given orally once daily for the same 56-day treatment period. The registry identifies the study model as parallel, so the comparison is between groups rather than a reported within-participant crossover.

The primary outcome assessment used the One Card Learning task, a visual-memory task based on a pattern-separation paradigm. The outcome was change from baseline through Week 8 in averaged correct-response rate. The primary analysis population was defined as participants who received at least one dose and had at least one measurement for that outcome. The posted outcome groups contain 23 donepezil participants and 13 placebo participants; those are the counts associated with this measure, not the registry's total enrollment.

The registry also lists two secondary outcomes: change from baseline in the standard deviation for averaged correct-response rate on OCL, using an arcsine square-root transformation, and change from baseline in averaged correct-response rate among participants receiving donepezil. The registry record does not report the allocation ratio, randomization procedure, masking details, or the original sample-size justification. It also does not report a detailed protocol for the repeated assessment schedule beyond describing the assessment as high-frequency and site-based.

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.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All participants who received at least one dose of study medication and yielded at least one measurement for that outcome measure.

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 Averaged Correct Response Rate on the One Card Learning Task to Week 8, measured over Baseline, Up to Week 8.

The primary endpoint is change from baseline in the averaged correct-response rate on the One Card Learning task at Week 8. OCL uses a pattern-separation paradigm to assess visual memory. The outcome is expressed as a proportion of correct responses. For each participant, the change is defined as the Week 8 averaged correct-response rate minus the baseline correct-response rate; the posted group summaries then describe the mean change and its standard deviation.

Using change from baseline puts the comparison on the scale of improvement or decline over the assessment period, rather than comparing only the final measurement. The donepezil-versus-placebo contrast asks whether the average change differs between the groups. The registry's primary hypothesis framed a difference of at least 2 percentage points in average correct responses as the target direction and magnitude. That stated hypothesis is distinct from the actual estimated contrast: the posted estimate is a mean difference of 0.021 in proportion units.

A computerized cognitive task can be useful in a study intended to assess whether repeated cognitive testing can capture potential treatment-related changes. The registry describes that assessment purpose, but does not report evidence about the task's reliability, validity, learning effects, or suitability for other populations. A numerical change on this endpoint is a measurement result; interpreting its clinical importance requires information that is not reported in the registry record.

7. Results Posted in the Registry

• Donepezil: mean 0.06 versus 0.03 on control; difference 0.02 (95% CI -0.04 to 0.08), standardized effect d = 0.25; z-test p = 0.484.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: each participant has a change-from-baseline value on the proportion scale, and the registry reports group means and standard deviations. The posted comparison is a mean difference between donepezil and placebo. The associated recalculation is an unadjusted comparison based on the posted summary statistics, rather than a participant-level model adjusted for covariates or other design features.

What does the mean difference represent?

The mean difference subtracts the placebo group mean change from the donepezil group mean change. A positive value means the observed average change was larger in the donepezil group; a negative value would mean it was larger in the placebo group. The posted value of 0.021 is in proportion units. It is an estimate of the contrast in these observed groups, not a guarantee that the same contrast would occur in another sample or setting.

The group means are 0.055 for donepezil and 0.034 for placebo. Their standard deviations are 0.078 and 0.091. Standard deviation describes how spread out participant-level changes are around a group mean. It is not the standard error of the mean and does not directly express the precision of the treatment contrast. Precision depends on the spread and the number of participants contributing to each group, among other assumptions.

What does the unadjusted calculation assume?

An unadjusted comparison uses the posted group summaries without controlling for baseline characteristics, stratification variables, repeated measurements, or other covariates. Such calculations can be useful for an independent check of a reported contrast, but they may not reproduce an analysis based on the trial's individual records or prespecified model. The registry record does not report the primary analysis model, any covariate adjustments, or whether the posted result arose from a different method. The recalculation should therefore be read as a simple summary-statistic comparison, not as a reconstruction of an undocumented full analysis.

For a continuous outcome, a mean difference retains the measurement scale, which makes its direction and size interpretable in the units of the endpoint. Here, those units are proportion of correct responses. A standardized effect can express a difference relative to variability, but no such standardized estimate is needed to interpret the mean difference and interval. The registry record does not report a standardized primary effect as its primary analysis result.

Why is a confidence interval useful?

A confidence interval gives a range of values that are reasonably compatible with the observed estimate under the calculation's assumptions. It helps distinguish the point estimate from the uncertainty around it. A narrow interval indicates more precision; a broad interval indicates less. The interval around this contrast includes both negative and positive values, so it does not isolate a single direction of difference. A confidence interval is not a probability statement that the true effect lies within the particular interval, and it does not replace consideration of the study design or analysis population.

The interval and the p-value address related but different questions. The interval displays a range of effect sizes compatible with the data and method. The unadjusted p-value, 0.484, quantifies how unusual data at least as discrepant as those observed would be under a specified no-difference model and its assumptions. It is not a measure of the probability that the treatment has no effect, nor does it show that a meaningful difference has been ruled out. The registry record does not state a threshold for interpreting the p-value or describe a multiplicity adjustment.

Why can variability affect precision?

When outcomes vary substantially among participants, the group average is less precisely estimated for a given group size. This is why standard deviations matter when considering the uncertainty around a mean contrast. If variability is greater than a method assumes, standard errors and intervals can be too small unless the analysis accounts for that extra spread. This general issue is called over-dispersion in settings where a model assumes less variability than the data show. The posted endpoint is a proportion-based continuous measure, and the registry does not report a model-based dispersion assessment or a correction for over-dispersion.

The posted standard deviations describe observed dispersion, but the registry does not supply participant-level distributions, diagnostic checks, or details about transformations used for the primary outcome. A secondary outcome does specify an arcsine square-root transformation for the standard deviation of averaged correct-response rate. That secondary description does not establish that the primary comparison used the same transformation.

How should the sponsor-reported result be read?

The registry reports a posterior probability value of 46.30 for the two groups, but does not identify the method or provide an interval or p-value for that analysis. Posterior probabilities depend on a statistical model and prior assumptions. Without those details, the value cannot be compared directly with a frequentist mean difference or p-value, and it cannot be translated into a treatment-effect estimate. The registry record does not explain whether the posterior analysis was the primary analysis, a supplementary analysis, or how it addressed the stated hypothesis.

The registry lists three outcome measures and three analyses, including the primary and secondary outcomes. It does not report how the analyses were ordered, whether any were prespecified as confirmatory, or how the chance of false-positive findings across multiple outcomes was handled. These omissions limit conclusions about the inferential status of any single result.

10. Confidence Intervals and Precision

The recalculated mean difference is 0.021, with donepezil minus placebo as the comparison direction. Its confidence interval extends below zero and above zero. Thus, the interval includes both a possible negative difference and a possible positive difference, as well as no difference. The point estimate alone may look more definite than the range of uncertainty supports.

A confidence interval represents uncertainty in the estimated group contrast under the calculation's assumptions. It is not a range containing every participant's outcome, and it is not a statement that there is a specified probability that the true effect falls within these particular limits. The group standard deviations, 0.078 for donepezil and 0.091 for placebo, describe variation among participant changes. They help distinguish outcome variability from uncertainty about the difference between group means.

The unadjusted p-value of 0.484 is consistent with substantial uncertainty about the direction and magnitude of the contrast. It does not establish that the groups have the same average change, and it does not rule out effects that could be important under a different interpretation of the endpoint. The registry does not report a prespecified equivalence margin or non-inferiority margin, so the result should not be interpreted as evidence of equivalence.

Precision is also limited by the number of participants contributing to the posted outcome groups: 23 in the donepezil group and 13 in the placebo group. The registry record does not report why these outcome counts differ from total enrollment, how missing endpoint values were handled, or whether the analysis used all available measurements in another way. Those details would help assess the possible influence of incomplete outcome data.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The sample-size figures are planning illustrations based on the observed effect, not the trial's original design calculation. The registry record does not report the original sample-size justification or design assumptions. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set at the stated fraction of the observed effect.

Every figure is participants per arm, not a total across both groups and not a total number of events. The table gives separate planning counts for 80% power and 90% power. In plain terms, the 80% column gives the participants per arm estimated to provide an 80% chance of detecting the chosen effect under the planning assumptions if that effect were the true one. The 90% column uses a higher probability of detection, so it requires a larger planned sample for the same effect. These are planning targets, not observed characteristics of the completed trial.

The rows consider planning for 100% of the observed effect, 75% of the observed effect, and 50% of the observed effect. For the full observed effect, the figures are 245 participants per arm for 80% power and 327 participants per arm for 90% power. For planning at 75% of the observed effect, they are 435 and 582 participants per arm, respectively. For planning at 50% of the observed effect, they are 977 and 1,308 participants per arm, respectively. Smaller effects require more participants because they are harder to distinguish from random variation under the same assumptions.

These are hypothetical design calculations based on the observed estimate, which may itself be imprecise. They are not observed power and should not be used to explain the result's p-value after the fact. A planning illustration can help show how assumptions about target effect and desired power affect sample size, but it does not establish what sample size the original investigators intended or whether a future study should use these exact figures.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = 0.021; pooled standard deviation σ = 0.083.
2. n = 2 × (1.96 + 0.842)² × 0.083² / 0.021² = 244.2, rounded up to 245 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
delta <- 0.021                      # observed difference in means
sigma <- 0.0828215658189694                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(4730635)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 245 (80% power), 327 (90% power)
 75% of observed effect: n per arm = 435 (80% power), 582 (90% power)
 50% of observed effect: n per arm = 977 (80% power), 1,308 (90% power)
power.t.test: 245.1 
Simulated power at n = 245 per arm: 0.801 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Analyses Reported by the Sponsor

The registry reports a posterior probability value of 46.30 for an analysis involving donepezil and placebo. It does not report the method, model, prior assumptions, interval, or a p-value for that analysis. The value therefore cannot be directly compared with the unadjusted recalculated mean difference of 0.021 or its confidence interval as though all were estimates on the same statistical scale.

The recalculation is a simple, unadjusted contrast derived from the posted group means, standard deviations, and group counts. The reported posterior probability is a separate result with insufficient methodological detail in the registry record to reproduce or interpret fully. The record does not identify whether it is the sponsor's primary inferential result, nor does it explain how the result relates to the stated primary hypothesis.

The discrepancy indicator is false, so no disagreement between a posted sponsor estimate and the recalculated estimate is identified in the record. Even so, the missing methodological details prevent a direct reconciliation of the posterior probability with the summary-statistic comparison. The two results should be kept distinct rather than treated as interchangeable evidence.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome and two secondary outcomes. The primary outcome is change from baseline in averaged correct-response rate on OCL at Week 8. The secondary outcomes concern change in the standard deviation of the averaged correct-response rate after an arcsine square-root transformation, and change in averaged correct-response rate among participants receiving donepezil. The record also lists three outcome measures and three analyses.

When multiple outcomes or analyses are examined, the chance of obtaining at least one apparently notable result can increase if each is interpreted independently. Statistical methods can address this issue through a prespecified hierarchy, an adjustment to significance thresholds, or other approaches. The registry record does not report whether any multiplicity-control strategy was planned or used, how the outcomes were prioritized beyond identifying the primary outcome, or whether analyses were confirmatory or exploratory. It is therefore not possible to determine from the record how the other analyses affect interpretation of the primary result.

The primary outcome population included participants who received at least one dose and yielded at least one measurement for the outcome. This is an explicitly stated analysis population, but the registry does not report the handling of participants who received treatment but lacked an endpoint measurement, the amount or pattern of missing data, or any imputation method. It also does not report whether a sensitivity analysis assessed the consequences of missing outcomes.

The posted outcome counts are 23 for donepezil and 13 for placebo, while total enrollment is 44. The registry record does not explain the difference between these counts or identify the reasons participants may not have contributed to the primary outcome. No further analysis population, such as an intention-to-treat population, is described in the record. The record also does not report the analysis model, covariate adjustment, or a detailed statistical analysis plan. These omissions make it difficult to assess how robust the posted comparison is to alternative analysis choices.

15. 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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: presyncope (3.4%).

The serious adverse event figures are counts among the groups identified in the registry. The donepezil group had 29 participants at risk for serious adverse events and 1 with a serious adverse event. The placebo group had 15 at risk and 0 with a serious adverse event. The registry records 0 deaths in each group. These counts describe what was reported for the study; they do not establish that an event was caused by or unrelated to study medication.

The registry lists Presyncope at 3.4% among the top serious adverse event terms. It does not provide further detail about the event in the information summarized in the record, so no additional interpretation about timing, severity, relationship, or recurrence is warranted. The record also lists other adverse events: 16 among 29 participants at risk in the donepezil group and 4 among 15 at risk in the placebo group. These are separate from the serious-event counts.

Adverse-event counts are descriptive and should be read alongside the number at risk in each group and the study's treatment period. The record does not provide enough information to assess comparative risk, event duration, or causal attribution. Small group counts also mean that the figures should not be treated as a precise estimate of event frequency beyond this study.

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

The posted results show a positive difference in average change on the OCL correct-response rate: 0.055 in the donepezil group compared with 0.034 in the placebo group, for an unadjusted mean difference of 0.021. This is the observed contrast for participants included in the posted outcome groups. The confidence interval includes both negative and positive differences, and the unadjusted p-value is 0.484. The estimate is therefore uncertain in both direction and magnitude.

The primary hypothesis described a difference of at least 2 percentage points in average correct responses. The reported mean difference is in proportion units and is not a substitute for a conclusion about clinical importance. The registry record does not establish what size of change would be meaningful to participants or clinicians, nor does it report an equivalence criterion. The result should not be interpreted as proof of benefit, proof of no effect, or evidence that the groups are equivalent.

The study's stated purpose was to assess whether repeated high-frequency, site-based computerized cognitive assessment could evaluate potential treatment effects. The posted endpoint provides an observed comparison on one cognitive task through Week 8. It does not, by itself, establish how well the platform measures cognitive change, whether the task can distinguish treatment effects in other studies, or whether findings generalize to a broader population.

Interpretation is further constrained by the outcome-group counts, the absence of reported missing-data methods, and the lack of a detailed analysis plan in the registry record. The separate posterior probability value is also not interpretable as a direct counterpart to the mean difference because its method is not reported. The evidence is best understood as a small, imprecise comparison reported for this study, rather than a definitive estimate of an effect or of the platform's performance in other settings.

17. Why This Trial Matters Statistically

This study raises a statistical question beyond the comparison of two group means: whether repeated computerized cognitive assessments can measure changes with enough consistency and precision to support evaluation of potential treatment effects. A cognitive outcome must capture meaningful differences while also accounting for ordinary variation in performance. The registry describes the assessment as high-frequency and site-based, but does not report detailed evidence about measurement reliability or the effect of repeated testing.

The posted primary estimate illustrates why effect size and precision need to be considered together. The mean difference is positive, but its confidence interval extends in both directions across no difference, and the p-value is 0.484. A point estimate gives one summary of the observed contrast; the interval shows that the data do not determine its direction and size precisely. This distinction is especially relevant when small group counts contribute to uncertainty.

The outcome is a proportion of correct responses, and the registry reports standard deviations for participant-level change. Understanding how variability, repeated assessments, and missing measurements were handled would help readers judge the analysis. The registry record does not describe those methods in detail. It also does not report the original sample-size rationale or the planned multiplicity strategy across the primary and secondary outcomes.

For biostatisticians and clinical researchers, the record is a useful example of separating an assessment's measurement aim from a treatment comparison, and of distinguishing a sponsor-reported posterior probability from a summary-statistic recalculation. Clear reporting of analysis populations, missing-data handling, model assumptions, and prespecified inferential priorities is central to making such results reproducible and appropriately interpretable.

18. 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 primary comparison is based on the participants reported for the outcome, 23 in the donepezil group and 13 in the placebo group, rather than the total enrollment of 44. The registry record does not explain why the counts differ from enrollment, whether outcome data were missing, or how any missing measurements were handled. The resulting uncertainty is reflected in a confidence interval that spans negative and positive differences.

The record does not report the original sample-size justification, a detailed statistical analysis plan, or the primary model. It does not state whether the comparison was adjusted for baseline measures or other variables. The simple recalculation is unadjusted and may not match a participant-level analysis. The reported posterior probability value cannot be fully interpreted because its method and assumptions are not described.

The registry identifies two secondary outcomes and three analyses but does not describe any multiplicity plan. It also does not report whether the primary outcome analysis was prespecified in the same form as the posted calculation, or how alternative analyses were considered. These gaps limit assessment of the inferential status and robustness of the reported comparison.

The study was a Phase 1 trial with 44 enrolled participants and a 56-day treatment period. The registry describes a specific population with mild cognitive impairment or mild Alzheimer's disease and an outcome on one computerized visual-memory task. It does not report evidence sufficient to assess the task's measurement properties, longer-term outcomes, or generalizability to populations beyond those enrolled. These limitations are specific to the information recorded and do not determine what may be available in other study documents.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Cognition Platform Study in Participants at Risk for Alzheimer's Disease (AD) (MK-0000-413) — NCT04730635.
• Registry record last updated 2024-10-28; status last verified 2024-07.
• Results first posted 2024-10-28.
• 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.

22. Record Summary

NCT04730635 is a completed, randomized, triple-masked, parallel-group Phase 1 study sponsored by Merck Sharp & Dohme LLC. The registry lists 44 enrolled participants with mild cognitive impairment or mild Alzheimer's disease. Participants received oral donepezil, at doses up to 10 mg once daily with scheduled titration, or oral placebo once daily for 56 days.

The primary outcome was change from baseline through Week 8 in averaged correct-response rate on the One Card Learning task. The posted group means were 0.055 for donepezil and 0.034 for placebo; the unadjusted mean difference was 0.021. Its confidence interval includes negative and positive differences, and the unadjusted p-value is 0.484. The registry also reports a posterior probability value of 46.30, without enough methodological detail to interpret it alongside the recalculation.

The registry record does not report the original sample-size justification, detailed missing-data methods, or a multiplicity strategy. The findings therefore describe an uncertain, short-term comparison on the recorded cognitive endpoint, with important analysis details not reported in the registry record.