COcoa Supplement and Multivitamin Outcomes Study for the Mind: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for COcoa Supplement and Multivitamin Outcomes Study for the Mind (NCT03035201), a registered phase not stated study. The primary outcome is Global Cognitive Function: Average Change Over Follow-up From Baseline.

2. Clinical Question

3. Trial Design

COSMOS-Mind is described as a cognitive ancillary study among men and women aged 65 years and older who were enrolled in the parent COcoa Supplement and Multivitamin Outcomes Study. The parent trial is described as randomized and as evaluating cocoa extract and a standard multivitamin supplement in relation to cardiovascular disease and cancer. That parent-trial description does not, by itself, establish the allocation procedure for this ancillary-study record: allocation is not stated in the registry record, and the listed arms are not a like-for-like control comparison as recorded.

The four listed arms represent combinations of cocoa extract or cocoa extract placebo with multivitamin or multivitamin placebo. The primary comparison groups participants by cocoa assignment across the multivitamin assignments: active cocoa flavanol, with or without active multivitamin, versus placebo cocoa flavanol, with or without active multivitamin. This is described as a comparison of the margins of the factorial design. The record does not report how assignment was generated or concealed, or whether other design features were used.

Enrollment is listed as 2262. The study start is listed as 2016-09, primary completion as 2021-02-12, and status as COMPLETED. The record says participants at the Wake Forest site will only be getting a phone call. It does not give further detail about the timing or conduct of those calls.

The primary endpoint is a global cognitive-function composite assessed from baseline through Year 3. The registry also lists secondary outcomes concerning composite cognitive function, mild cognitive impairment, executive function and episodic memory. The record does not give the results for those outcomes. It also does not report the protocol, recruitment flow, adherence, or numbers analyzed for each outcome.

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: Intention to treat -- All enrolled participants

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 Global Cognitive Function: Average Change Over Follow-up From Baseline, measured over From Baseline through Year 3.

The primary endpoint is the average change in global cognitive function from baseline through Year 3, expressed as a score on a scale. It combines individual test z-scores from the Telephone Interview for Cognitive Status-modified Total and Long, Immediate and Delayed Story Recall, Oral Trail Making A and B, Category Fluency, Verbal Fluency, and Number Span Backward and Forward. The Oral Trail Making scores are log-transformed. Scores are oriented so that higher positive values indicate better performance.

The primary contrast compares the average of the composite scores at Years 1, 2 and 3 with the baseline composite score. In general, combining multiple tests into a composite can represent performance across several cognitive tasks rather than focusing on one test alone. A longitudinal contrast also uses measurements at multiple follow-up times to characterize change over the stated period. The record specifies this contrast and says it is fitted using a random effects model for repeated measures.

The endpoint is relevant to the study question because the ancillary study is intended to examine cognitive function and risk of cognitive impairment. However, the record does not report a numerical estimate of the primary contrast, its standard error, confidence interval, or test result. The endpoint definition therefore describes what was planned for measurement, not what the observed data showed.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry record describes the primary analysis as a comparison of cocoa-flavanol assignment groups across the multivitamin assignments. It also names a random effects model for repeated measures. The explanation below distinguishes those recorded features from general statistical concepts; details not stated in the record remain unknown.

What does the primary contrast represent?

The primary outcome is a change contrast: the average composite score at Years 1, 2 and 3 minus the baseline composite score. In general, a contrast is a defined combination of measurements that expresses the question an analysis is intended to answer. Here, the contrast summarizes follow-up performance relative to baseline over the stated period. It does not reduce the endpoint to a single test, because the composite draws on multiple cognitive assessments.

The composite is built from individual test z-scores. In general, a z-score expresses a measurement relative to a reference distribution, in standardized units. The registry record does not describe the reference distribution or additional scoring details. It does state that the tests are combined and that higher positive composite scores correspond to better performance. Oral Trail Making A and B scores are log-transformed before contributing to the composite, according to the endpoint description.

Why use a repeated-measures model?

Repeated measurements from the same participant are generally correlated: a person's scores over time are not independent observations. A random effects model can represent participant-level variation while modeling repeated outcomes. In general, accounting for within-participant structure helps avoid treating repeated scores as though they came from unrelated people. The record says that the specified linear contrast is fitted using a random effects model for repeated measures, but it does not report the model's covariance structure, fixed effects, estimation method, or handling of incomplete follow-up.

The primary result would ordinarily be interpreted using an estimated contrast and a measure of uncertainty. A standard error describes the estimated variability of an estimate across repeated samples under the model. A confidence interval gives a range of values compatible with the data and model at a stated confidence level. Neither the primary estimate nor its uncertainty is reported in the registry record, so the precision of the result cannot be assessed from the recorded endpoint description.

What is the factorial comparison?

The listed treatment combinations have two components: cocoa extract or cocoa extract placebo, and multivitamin or multivitamin placebo. The primary comparison is described as active cocoa flavanol, with or without active multivitamin, versus placebo flavanol, with or without active multivitamin. In general, a factorial comparison across a component's margins combines information across levels of another component. This makes the stated estimand a cocoa-flavanol comparison averaged across the multivitamin assignments, rather than a direct comparison of only two individual combination arms.

The registry record does not state allocation for this ancillary record. Although the parent-trial summary calls COSMOS randomized, the record's own arm descriptions do not document the assignment method for COSMOS-Mind. Consequently, the method used to form groups, and any consequences for causal interpretation, cannot be determined from the record alone.

How should planning calculations be read?

In general, a sample-size planning calculation estimates how many participants may be needed under specified assumptions to detect a chosen effect with a chosen error rate and power. The planning assumptions described for the illustrations are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The effect to detect is specified as a stated fraction of the observed effect. The record says that non-time-to-event illustrations give participants per arm; time-to-event illustrations give the total number of events.

Such calculations are planning illustrations, not the study's observed power and not a replacement for the trial's design rationale. The record does not report the protocol's sample-size justification or the numeric planning outputs in the information available for this explanation. It also does not report the observed primary effect needed to assess those illustrations. The assumptions should therefore not be mistaken for a complete description of the original study design.

What analysis details remain unspecified?

The record names an intention-to-treat population of all enrolled participants and describes the primary model and contrast. It does not report the statistical analysis plan, treatment of missing values, any interim analyses, adjustment for multiple comparisons, or whether any covariates were included. It also does not provide model diagnostics or sensitivity analyses. These omissions limit the ability to assess the robustness and reproducibility of the analysis from the registry record alone.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Cocoa Extract + Multivitamin", "Cocoa Extract + Multivitamin Placebo", "Cocoa Extract Placebo + Multivitamin", "Cocoa Extract Placebo + Multivitamin Placebo")
n   <- c(571, 553, 551, 587)
m   <- c(0.24, 0.15, 0.2, 0.15)
sp  <- c(0.02, 0.02, 0.02, 0.02)
sd  <- sp * sqrt(n)      # standard errors converted to SDs
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Cocoa Extract + Multivitamin  571  0.240  0.478  0.201  0.279
CARM  Cocoa Extract + Multivitamin Placebo  553  0.150  0.470  0.111  0.189
CARM  Cocoa Extract Placebo + Multivitamin  551  0.200  0.469  0.161  0.239
CARM  Cocoa Extract Placebo + Multivitamin Placebo  587  0.150  0.485  0.111  0.189

11. Multiplicity, Missing Data and Analysis Populations

The record lists a primary endpoint and several secondary outcomes, including composite cognitive function, incident mild cognitive impairment, mild cognitive impairment among those free at baseline, executive function and episodic memory. Some outcome labels appear more than once in the listed set. The record reports 8 outcome measures and 8 analyses, but does not describe how those analyses map to the listed outcomes or how testing across them was handled.

When many outcomes or comparisons are tested, the chance of at least one apparently unusual result can increase. In general, multiplicity procedures can control that risk, or an analysis plan can distinguish a primary question from exploratory questions. The record does not report a multiplicity adjustment, a hierarchy among secondary outcomes, or whether secondary analyses were intended to be confirmatory or exploratory. The statistical analysis plan is not reported in the registry record.

The primary analysis population is described as intention to treat, all enrolled participants. The record does not report the number with observed primary endpoint data, the extent or pattern of missing follow-up measurements, or how missing data were handled. It does not say whether participants with incomplete repeated measures were included through the model, whether imputation was used, or whether sensitivity analyses assessed missing-data assumptions. Those approaches cannot be inferred from the mention of a repeated-measures model alone.

The record also does not report analysis-population counts by arm, exclusions after enrollment, protocol deviations, or adherence. These details matter for understanding which observations contributed to each analysis. In the absence of an outcome estimate and a fuller analysis description, neither multiplicity control nor the consequences of missing data can be evaluated from the registry record.

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

The registry record defines a cognitive outcome and a planned comparison, but it does not report the primary result. It therefore does not establish the size, direction or precision of any difference in global cognitive-function change. The absence of a reported estimate is not evidence that the groups had the same outcome, and it is not evidence of a difference.

The endpoint concerns a composite formed from several cognitive tests, measured at baseline and through Year 3. Its interpretation is specific to that score and contrast. The record does not provide the underlying test results, an estimate for any individual test, or the numerical composite change. It also does not report the listed secondary outcome results for mild cognitive impairment, executive function or episodic memory.

The primary comparison is across the cocoa-flavanol margins of the factorial design, combining participants with and without active multivitamin assignment within each cocoa group. It should not be read as the effect of the combination of active cocoa extract and active multivitamin alone. The record does not state allocation for this ancillary-study comparison, so the listed arm structure by itself does not establish a randomized contrast in this record.

In general, an estimated difference should be interpreted alongside its uncertainty, study conduct and analysis assumptions. Here, no primary estimate, confidence interval or p-value is reported in the registry record, and key details about missing data, multiplicity and the analysis plan are also absent. The completed status and the description of the intended endpoint do not fill those gaps. The record supports a description of the research question and analytic approach, but not a conclusion about whether supplementation changed cognitive outcomes.

13. Why This Trial Matters Statistically

This record is statistically useful because it makes the primary estimand more specific than a broad question about supplements and cognition. It describes a composite outcome, the timing of measurement, the contrast from baseline, and a comparison across the cocoa-flavanol margins of a factorial design. Those details help distinguish the target question from a simple comparison of one treatment combination with the all-placebo combination.

The endpoint also illustrates choices involved in constructing a cognitive outcome. It combines several tests, applies a log transformation to two Oral Trail Making measures, and orients the score so that higher positive values correspond to better performance. Such choices define what the resulting score means and what dimensions of performance it represents. A composite can make a broad outcome possible, but it does not reveal how each component test behaved; the individual results are not reported in the registry record.

The named repeated-measures model reflects the longitudinal structure of the planned analysis. In general, repeated measurements from a participant are related, and statistical methods need to account for that dependence. The registry description identifies the general modeling approach but does not provide enough detail to reproduce it or assess its assumptions.

The record also shows why registration details and posted results need to be distinguished. It identifies enrollment, study status, endpoint definitions and the intended analysis population, yet does not report the primary effect estimate or its uncertainty. Statistical interpretation depends on those outcome results and on details of missingness, multiplicity and the analysis plan. Without them, the record is informative about the question and design structure but does not support a conclusion about the observed cognitive effect.

14. 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 main limitation for interpreting this record is that it does not report the primary outcome estimate or its uncertainty. There is no numerical contrast, standard error, confidence interval or p-value in the registry record. Completion status and the stated analysis method cannot substitute for those results.

Allocation for this ancillary-study record is not stated. The parent-trial summary describes COSMOS as randomized, but the record does not document the allocation method for COSMOS-Mind. Its listed arms therefore do not, by themselves, establish a like-for-like control comparison or clarify the basis for causal interpretation.

The endpoint description is detailed about the composite components and contrast, but the record does not report how the composite was standardized, the model's covariance structure, or other implementation details. It does not describe missing-data handling, analysis deviations, sensitivity analyses, or multiplicity procedures. The listed secondary outcomes also lack numerical results in the registry record.

The record describes the analysis population as all enrolled participants under an intention-to-treat approach, but does not give endpoint-specific counts or the number of participants with follow-up data. It provides no adherence information or details about treatment exposure. It also does not report the protocol's sample-size justification. These gaps limit assessment of precision, robustness and the relationship between the intended and conducted analyses.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: COcoa Supplement and Multivitamin Outcomes Study for the Mind — NCT03035201.
• Registry record last updated 2023-08-22; status last verified 2023-05.
• Results first posted 2023-08-22.
• 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.

18. Record Summary

The registry record for the COcoa Supplement and Multivitamin Outcomes Study for the Mind lists Wake Forest University Health Sciences as sponsor, enrollment of 2262, and status COMPLETED. The study start is listed as 2016-09 and primary completion as 2021-02-12. It concerns men and women aged 65 years and older enrolled in the parent COSMOS study; the record says participants at the Wake Forest site will only be getting a phone call.

The primary endpoint is average change in a global cognitive-function composite from baseline through Year 3. The planned comparison is active cocoa flavanol versus placebo flavanol across the multivitamin assignments, using a random effects model for repeated measures. The primary analysis population is described as intention to treat, all enrolled participants.

The registry record does not report the primary outcome estimate or its precision. Allocation for this record, missing-data methods, multiplicity handling and the statistical analysis plan are not reported in the registry record.