Music & MEmory: A Pragmatic TRial for Nursing Home Residents With ALzheimer's Disease (METRIcAL)_part2: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Music & MEmory: A Pragmatic TRial for Nursing Home Residents With ALzheimer's Disease (METRIcAL)_part2 (NCT04850807), a randomized na study. The primary outcome is Frequency of Agitated and Aggressive Behaviors (Staff Report).

2. Clinical Question

3. Trial Design

This was a randomized, parallel-group trial with single masking. The registry records enrollment of 990 and a completed status. The recorded start date is 2021-05-14, and the primary completion date is 2022-09-08. The phase is listed as NA. Brown University is the sponsor.

The design is cluster-randomized. That feature matters when interpreting the statistical results: assignment is organized through clusters rather than treated as independent assignment of every resident. The registry record does not describe the cluster unit, the number or size of clusters, or the exact allocation procedure beyond identifying the trial as randomized. It also does not specify who was masked. Single masking is recorded, but the masked role or roles are not reported in the registry record.

The intervention arm received Music & Memory, a program in which staff offered personally tailored playlists at early signs of agitation. The comparator arm received usual care for managing agitated and/or aggressive behaviors in the nursing-home setting. The brief description frames the trial as testing the program’s effects on those behaviors among nursing-home residents with dementia. That description states the research purpose; it is not, by itself, evidence of an effect.

The primary outcome was frequency of agitated and aggressive behaviors by staff report, assessed with the CMAI at baseline and follow-up. Its stated timeframe is 4-months. Secondary outcomes listed in the record include administrative data on behavior frequency, use of antipsychotic, antianxietal, antidepressant, and hypnotic medications, and depression. The record lists 7 outcome measures and 6 analyses. It does not report the detailed analysis population, handling of missing outcome measurements, or a fuller statistical analysis plan. Those omissions limit how completely the design can be reconstructed from the registry record.

4. Participant Flow and Accounting

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

Overall Study

Reasons for not completing this period:

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 Frequency of Agitated and Aggressive Behaviors (Staff Report), measured over 4-months.

The primary endpoint concerns how often residents exhibited agitated or aggressive behaviors, as reported by staff who knew the resident well. Research staff interviewed nursing staff using the Cohen-Mansfield Agitation Inventory. The instrument asks about 29 behaviors during the past week. Each item has seven response choices, anchored from never (1) to several times per hour (7). The total score ranges from 29 to 203, with higher scores representing more frequent behaviors.

The CMAI was administered at baseline and follow-up, and the primary outcome timeframe is recorded as 4-months. Measuring behavior through staff interviews gives the trial a structured way to assess changes in behaviors that staff may observe in the nursing-home setting. A total score also provides a continuous measure, which can represent variation across the scale rather than reducing the outcome to whether a participant did or did not exhibit a behavior.

The endpoint is a score on a scale, and the result is expressed as an average marginal effect. The reported estimate is -0.4, with a 95% confidence interval from -6.9 to 6.0. Since the registry record does not state the order in which the arms were contrasted, the negative sign cannot by itself identify which arm had the lower estimated score. Nor does the record provide a threshold for what size of score change would be important to residents or staff. The endpoint therefore describes reported behavior frequency, while the practical meaning of a particular score difference is not established by the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a Differences-in-Differences analysis compare?

Differences-in-Differences is a general statistical approach that compares change over time between groups. Rather than interpreting only the follow-up difference, the method uses the before-and-after pattern in each group and contrasts those changes. In this trial, the primary measure was collected at baseline and follow-up, which makes a change-based comparison relevant to the stated outcome. The registry identifies the sponsor’s method as Differences-in-Differences regression and reports its result as an average marginal effect.

In general, a regression model can account for measured features of the data and express a group contrast in terms of predicted outcomes or marginal effects. An average marginal effect summarizes an average modeled contrast on the outcome scale. Here, that scale is the CMAI score. The registry record does not report the model formula, the included covariates, the precise marginal-effect calculation, or the assumptions used to interpret the change comparison. Those details should not be inferred from the method name alone.

Why does cluster randomization matter?

In a cluster-randomized design, groups or settings, rather than individuals considered independently, are assigned to trial arms. People within the same cluster may have more similar outcomes than people in different clusters. General statistical methods account for that dependence because treating all participants as independent can make the apparent amount of information too large and uncertainty too small. The registry explicitly warns that a simple individual-level re-analysis would be misleading and therefore reports only the sponsor’s own analysis.

The record does not identify the cluster unit, the number of clusters, cluster sizes, or the specific method used to account for within-cluster dependence. Those omissions mean that readers cannot reconstruct the cluster-adjusted uncertainty from the registry entries alone. They also make it inappropriate to use the enrollment figure as if it were a complete description of the independent information available to the analysis.

How should the confidence interval and p-value be read?

A confidence interval is an uncertainty interval generated by a statistical procedure. In general, a 95% confidence interval describes the performance of the procedure over repeated applications under its assumptions; it is not a guarantee that the true effect has a 95% chance of falling within this particular interval. The reported interval, -6.9 to 6.0, includes zero and covers contrasts in either direction. Its span indicates that the estimate is not precise enough to pin down a narrowly defined effect from the posted result alone.

A p-value summarizes how unusual data at least as incompatible with a specified null model would be, assuming that model and the analysis assumptions. It is not the probability that the null hypothesis is true, and it does not measure the size or practical importance of an effect. The posted p-value is <0.05, while the posted 95% interval includes zero. For conventional matching two-sided procedures, those statements would usually point to different conclusions about whether zero is excluded at the corresponding threshold. The registry record does not report why they differ, so readers should not treat the discrepancy as resolved.

The contrast order is not identified in the registry record. Even though higher CMAI scores correspond to more frequent behaviors, interpreting whether a negative average marginal effect favors Music & Memory or usual care requires knowing which arm is subtracted from which. That information is not reported in the record. A numerical sign without a defined contrast order is not enough to assign a direction to the treatment comparison.

What do planning calculations represent?

Planning illustrations based on the observed effect are not measures of the trial’s observed power. The registry describes their assumptions as a two-sided significance level of 0.05, equal allocation between the arms, and a normal approximation, with the effect to detect expressed as a stated fraction of the observed effect. Such calculations illustrate planning quantities under chosen assumptions; they do not establish the probability that the completed trial detected an effect, and they are not the trial’s own design justification.

The record does not report the trial’s sample-size justification or enough information to verify whether the planning assumptions match the original design. For a cluster-randomized study, planning also depends on features of clustering and outcome variation. The details needed to assess those features are not reported in the registry record. Any planning illustration should therefore remain distinct from the sponsor’s primary estimate and its uncertainty.

What can and cannot be evaluated from the reported method?

The method name and result identify the broad analysis strategy, but they do not constitute a full statistical analysis plan. The registry record does not report the exact model specification, adjustment variables, handling of missing data, analysis population, or whether any sensitivity analyses were conducted. It does not report an explanation for the p-value and confidence-interval tension. These are not grounds for inventing alternate analyses; they mark limits on what can be assessed from the recorded details.

The registry lists multiple outcomes, including behavior frequency from administrative data, several medication-use outcomes, and depression. The presence of multiple outcomes raises a general question about multiplicity: when many hypotheses are assessed, the chance of at least one apparently unusual result can increase. Whether and how this trial addressed that issue is not reported in the registry record. The primary result should be kept distinct from the secondary outcome analyses, and the record does not justify assuming a correction or a prespecified hierarchy.

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("Music & Memory", "Usual Care")
n   <- c(431, 419)
m   <- c(44.1, 47.9)
sp  <- c(3.1, 3.1)
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  Music & Memory  431  44.100  64.358  38.007  50.193
CARM  Usual Care  419  47.900  63.455  41.806  53.994

11. Analyses Reported by the Sponsor

The posted primary result is the sponsor’s own Differences-in-Differences regression analysis, reported as an average marginal effect of -0.4 with a 95% confidence interval from -6.9 to 6.0 and p <0.05. The registry record does not present a simple unadjusted re-calculation of this comparison. It identifies the study as cluster-randomized and states that a simple individual-level re-analysis would be misleading; only the sponsor’s own analysis is shown.

That distinction matters because a calculation from participant-level summary statistics would not necessarily preserve the cluster structure or the sponsor model’s treatment of dependence. The reported estimate should therefore be read as a model-based sponsor result, not as a quantity independently verified from arm-level summaries. The record does not provide the model details needed to reproduce the regression or assess all of its assumptions.

The posted p-value and interval also warrant careful reading together. The interval includes zero, while the p-value is reported as <0.05. The registry does not explain the apparent lack of agreement under the usual relationship between a two-sided confidence interval and test. No alternative result should be substituted for the sponsor’s analysis, and the record does not establish which inferential detail accounts for the difference. The safest interpretation is to preserve both reported statistics and recognize that their relationship remains unexplained.

12. Multiplicity, Missing Data and Analysis Populations

The registry lists 7 outcome measures and 6 analyses. Alongside the primary staff-reported behavior endpoint, the secondary outcomes include frequency of agitated and aggressive behaviors from administrative data, antipsychotic use, antianxietal use, antidepressant use, hypnotic use, and depression. These outcomes cover distinct data sources and clinical domains, so they should not be treated as interchangeable measures of the same endpoint.

When several outcomes or hypotheses are assessed, multiplicity is a general statistical concern: the chance of obtaining at least one nominally unusual result can rise as the number of tests grows. Methods such as an adjustment, a prespecified testing hierarchy, or a distinction between confirmatory and exploratory analyses can address related concerns. Whether any such approach was planned or used is not reported in the registry record. The record also does not explain why it lists 7 outcome measures and 6 analyses, so no one-to-one mapping between those counts should be assumed.

The missing-data approach is not reported in the registry record. The record does not say how incomplete CMAI interviews, unavailable administrative data, or missing medication or depression measurements were handled. It also does not describe whether any imputation, complete-case analysis, or sensitivity analysis was used. Each choice can affect the population contributing to an estimate and the interpretation of uncertainty, but none should be guessed.

The analysis population is likewise not reported in the registry record. The entry does not state whether the primary analysis followed the randomized assignment regardless of intervention uptake, used another defined population, or applied any exclusions after randomization. It does not report how cluster membership, transfers, or other changes during follow-up were handled. These absences prevent a full assessment of the estimand and the potential influence of missing observations.

Consequently, the primary sponsor result can be described as posted, but it cannot be placed within a fully documented multiplicity, missing-data, and analysis-population framework. A lack of registry detail does not demonstrate that these issues were ignored; it means the record does not state how they were addressed.

13. 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.

The registry reports serious adverse events affecting 0 participants at risk in Usual Care, with 489 at risk, and 0 participants at risk in Music & Memory, with 501 at risk. It also reports 0 other adverse events in each arm, with 489 and 501 at risk, respectively. These are the posted counts and denominators; they should not be converted into comparative rates without an appropriate analysis.

Deaths are reported separately as 50 in Usual Care and 55 in Music & Memory. The record does not state whether these deaths were considered related to the intervention, how they were classified in relation to adverse-event reporting, or how timing and follow-up were handled. A count alone cannot establish cause or a comparative safety effect. No serious-event terms are listed in the registry record.

These figures describe the posted safety reporting, not a complete interpretation of risk. The record does not give additional detail on event definitions, severity grading, exposure duration, adjudication, or the relationship between the death counts and the adverse-event categories. Zero reported affected participants in an adverse-event category should not be read as proof that no event of any kind occurred; it is a statement about the counts recorded in those categories.

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

The primary result is an uncertain estimate for a staff-reported measure of agitated and aggressive behaviors. The point estimate is -0.4, and the 95% confidence interval extends from -6.9 to 6.0. Because the interval includes zero and spans both negative and positive values, the posted result does not precisely determine the direction or size of the contrast. The registry record also does not identify which arm is subtracted from the other, so the sign cannot be translated into a treatment-favoring direction.

The reported p-value is <0.05, but the interval includes zero. Since the registry record does not explain the tension between those summaries, neither should be omitted when describing the result. The p-value alone does not describe the magnitude or practical importance of an effect, and the interval does not establish whether any particular value would matter to residents or staff. A clinically meaningful score threshold is not reported in the registry record.

The result does not show that the personalized music program works or does not work. It reports the sponsor’s statistical comparison for the designated primary endpoint at the 4-months timeframe. It does not settle the secondary questions about administrative behavior records, medication use, or depression. Nor does it answer questions about uptake, fidelity, or longer-term patterns, which are not described in the registry record.

Interpretation is further bounded by the cluster-randomized design and the lack of detailed model and analysis information. The sponsor’s own regression is the analysis posted, because a simple individual-level re-analysis would be misleading. The registry does not give enough detail to reproduce the analysis or to evaluate all assumptions. These limits call for a careful reading of the reported estimate rather than an unsupported conclusion about benefit or harm.

15. Why This Trial Matters Statistically

Statistically, this trial is notable for combining a behavior-focused primary outcome with a cluster-randomized design. The CMAI provides a continuous score based on staff reports of 29 behaviors, and higher total scores indicate more frequent agitated or aggressive behaviors. Its use at baseline and follow-up makes change over time relevant to the primary comparison. The reported Differences-in-Differences regression is consistent with a change-based analysis, while cluster assignment makes it important to account for dependence among participants within clusters.

The registry’s warning against a simple individual-level re-analysis highlights a broader principle: the unit of randomization and the structure of the data matter for uncertainty estimates. A result cannot be independently checked by treating each resident as an isolated randomized unit if assignment occurred by cluster. The record’s limited description of clusters and modeling, however, leaves readers unable to reconstruct how that dependence was handled.

The reporting also illustrates why an effect estimate, confidence interval, and p-value should be considered together. The estimate is -0.4, the interval ranges from -6.9 to 6.0, and the p-value is <0.05. The interval includes zero, creating an apparent tension with the reported p-value that the record does not explain. This makes transparent reporting of inferential procedures especially important: readers need enough detail to understand whether the summaries arise from matching methods and what contrast is being estimated.

Finally, the record lists several secondary outcomes, while multiplicity handling, missing-data methods, and analysis populations are not described. That does not demonstrate that the study lacked such plans, but it restricts what can be concluded from the registry alone. Distinguishing a recorded result from a fully documented analysis is an important part of statistical interpretation.

16. 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.

Several limitations specific to the registry record constrain interpretation. First, the trial is cluster-randomized, but the cluster unit, number and size of clusters, allocation details, and method for accounting for within-cluster dependence are not reported in the registry record. This prevents readers from reconstructing the effective information structure or independently assessing the uncertainty calculation. The record explicitly cautions that a simple individual-level re-analysis would be misleading.

Second, the sponsor’s primary result is reported with a 95% confidence interval from -6.9 to 6.0 and p <0.05. The interval includes zero, and the registry does not explain the apparent tension between these summaries. The order of the arm contrast is also not reported, limiting interpretation of the estimate’s sign despite the CMAI’s scale direction. A clinically meaningful score difference is not reported either.

Third, the registry does not report the detailed statistical model, adjustment variables, analysis population, missing-data handling, multiplicity approach, or sensitivity analyses. It does not state who was masked, although masking is recorded as single. It also does not provide further detail about usual care, intervention uptake, adherence, implementation fidelity, or the extent to which the program was delivered as intended. These are omissions in the registry record, not evidence that the trial did not address them elsewhere.

Finally, the outcome timeframe is listed as 4-months, and the record does not establish longer-term effects. The safety entry reports adverse-event counts and deaths but does not describe attribution or event timing. The trial’s listed secondary outcomes are broader than the primary staff-reported score, but their results and multiplicity context are not available in the registry record described for this analysis. Accordingly, conclusions should stay limited to the posted primary estimate and the recorded safety counts.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: Music & MEmory: A Pragmatic TRial for Nursing Home Residents With ALzheimer's Disease (METRIcAL)_part2 — NCT04850807.
• Registry record last updated 2025-02-28; status last verified 2025-02.
• Results first posted 2025-02-28.
• 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.

20. Record Summary

NCT04850807 is a completed randomized, single-masked, parallel-group, cluster-randomized trial sponsored by Brown University. Enrollment was 990. The recorded start date is 2021-05-14, and primary completion is dated 2022-09-08. The study compared the Music & Memory personalized playlist program with usual care for nursing-home residents with dementia.

The primary endpoint was staff-reported frequency of agitated and aggressive behaviors on the CMAI at the 4-months timeframe. The sponsor’s Differences-in-Differences regression reported an average marginal effect of -0.4, a 95% confidence interval from -6.9 to 6.0, and p <0.05. The registry record does not explain the apparent tension between the interval and p-value or identify the contrast order. It also does not report detailed model specifications, missing-data handling, analysis populations, or multiplicity methods. Safety reporting lists 0 serious and 0 other adverse events in both arms, alongside deaths reported as 50 and 55, respectively.