Sleep and Circadian Treatments for Shift Workers: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Sleep and Circadian Treatments for Shift Workers (NCT03813654), a randomized na study. The primary outcome is Total Sleep Time (TST).

2. Clinical Question

The study used three components: a field trial, a shift worker survey, and focus groups. The survey was intended to examine demographic and operational factors related to adopting the sleep timing strategy. The focus groups were intended to explore reasons older workers might be unable or unwilling to adopt it. These components provide context for the field trial question, but their findings are not reported in the registry record.

The field-trial comparison is about sleep duration during the sleep episode or episodes between shifts, not simply time in bed. The primary outcome was derived from actigraphy. The record does not report whether the groups differed on other outcomes in a way that would explain or qualify the sleep-time summaries.

3. Trial Design

The registry describes a randomized, parallel-group study with no masking. It lists an enrollment of 33 and identifies Brigham and Women's Hospital as the sponsor. The study began on 2020-08-20, had a primary completion date of 2023-10-01, and is listed as terminated. The reason for termination is not reported in the registry record.

Three groups are represented in the primary outcome results: control, 8-h afternoon-evening sleep, and 8-h free sleep. In the control condition, participants were not given instructions about when or how long to sleep. The afternoon-evening group was directed toward a specified 8-hour sleep opportunity. The free-sleep group was instructed to remain in bed for 8 continuous hours but could choose the timing. The interventions were behavioral.

The field-trial intervention block followed the first night shift. Participants were told their randomization group at the end of that shift. The assigned instructions applied before the next two night shifts. The primary outcome timeframe, however, describes total sleep time averaged for up to 3 intervention nights. The registry does not explain how the available nights varied among participants or how the average was handled when fewer nights were available.

The overall study included more than the field trial: the survey and focus groups were designed to examine adoption and acceptability. The brief summary describes an aim to minimize sleep deficiency and related sleepiness and performance impairments, but the posted primary endpoint is sleep time. The registry lists 15 outcome measures and reports 4 secondary outcomes, including fatigue and inter-shift recovery, subjective stress, quality of life, and salivary cortisol. The cortisol aim was dropped when the study went fully remote due to COVID.

Details about the randomization sequence, allocation concealment, the number assigned to each group, and any planned interim analyses are not reported in the registry record. The listed arm counts for the primary outcome refer to participants included in those outcome summaries; they should not be read as a complete account of allocation or enrollment.

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: Enrollment was assigned upon start of the Baseline period, but participants were not informed of their randomization until after the first night shift of the Intervention period.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All participants with useable actigraphy data were included.

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 Total Sleep Time (TST), measured over The TST averaged for up to 3 intervention nights (in minutes) will be compared between groups..

Total sleep time (TST) is the primary endpoint. It measures minutes of sleep in the main sleep episode between shifts, based on actigraphy recorded in 60-second epochs. The registry says a commercial software package was used to score sleep. For control participants who slept ad lib, the primary description allows the major sleep episode to be used and also allows sleep across episodes to be summed when sleep occurred in more than one bout.

The endpoint is expressed as a mean in minutes, with standard deviation as the measure of spread. The planned summary was TST averaged for up to 3 intervention nights and compared between groups. Measuring sleep duration in the interval between shifts is relevant to the stated field-trial aim of examining sleep deficiency during night work. Actigraphy provides a recorded estimate of sleep and wake over time; it is not a direct measure of every aspect of sleep quality, alertness, or work performance.

As a general statistical point, a group mean describes the average recorded value, while a standard deviation describes how much individual observations vary around that mean. Neither alone describes all features of the data. The registry does not report the scoring software, actigraphy processing decisions, or handling of borderline or invalid recordings in enough detail to assess those procedures fully.

The endpoint captures sleep duration, not whether participants followed the assigned timing instructions, found them acceptable, or experienced improved performance. Those are separate questions. The registry record does not report primary-outcome estimates for each individual intervention night or a measure of adherence to the assigned sleep opportunity.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary measure is continuous: TST is recorded in minutes, and the registry identifies the reported summary as a mean with standard deviation. This makes the outcome interpretable on its original scale. For example, a mean is the average observed sleep time in a group; the standard deviation describes the dispersion of individual values around that average. Neither measure reveals the full shape of the data, and the registry does not report distributions, medians, or other summaries.

What does a mixed model do?

As a general statistical explanation, mixed models can account for data that have a grouped or repeated-measure structure. They can estimate group differences while representing dependence among observations, such as repeated measurements from the same participant. The registry reports a mixed models analysis, and the endpoint timeframe refers to TST averaged across up to 3 intervention nights. However, the model specification is not reported in the registry record: the fixed and random effects, covariance structure, any covariate adjustment, and the exact observations included are not described. It is therefore not possible to determine from the record how repeated nights were handled or what assumptions the fitted model used.

How should the reported mean difference be read?

The registry reports a mean difference (final values) of 90 for the mixed models analysis and p < 0.05. It does not name a specific pairwise contrast or explain whether the value summarizes an overall group comparison or another model contrast. The number should not be interpreted as a particular difference between two named arms without that information. The registry also does not report a confidence interval for this result. A confidence interval, as a general concept, communicates uncertainty around an estimated effect; its absence limits assessment of the estimate's precision and plausible range.

The reported p-value is below the registry's stated threshold in the reported analysis, but it does not indicate the probability that a hypothesis is true. Nor does it convey clinical importance, demonstrate adherence, or resolve the size of every group contrast. The precise meaning depends on the model, contrast, and testing procedure, which are not fully described in the registry record.

Why are descriptive summaries not the same as the model result?

Group means and standard deviations summarize the observed outcome values. A model-based result may instead use a defined contrast and account for repeated observations or other modeled features. Those approaches can answer related but different questions. Reconstructing a contrast from arm summaries alone would require a clearly specified comparison and suitable information about the data structure. The registry characterization states that values are missing or not estimable for a two-arm comparison of this continuous endpoint, so no such recalculation is presented.

As a general statistical point, over-dispersion means that observations vary more than a model's assumed distribution allows. It can make standard errors too small and tests too optimistic if the model does not account for the extra variation. The posted endpoint is continuous and reported with means and standard deviations; the registry does not describe an over-dispersion assessment or indicate that this issue was part of its analysis. The concept is relevant generally, but no conclusion about over-dispersion in these data can be drawn from the record.

What can the group spreads tell readers?

The reported standard deviations are 68.7 for control, 53.2 for 8-h afternoon-evening sleep, and 66.7 for 8-h free sleep. They show the scale of variation in TST within each reported group. They do not by themselves indicate whether a difference between group means is statistically reliable. That depends on the comparison, the sample information, the model, and the uncertainty calculation. No confidence intervals for arm means or pairwise contrasts are reported.

General explanation: a smaller standard deviation indicates less dispersion around a group's mean in the observed data, but it does not necessarily mean the estimate of the group mean is more precise in every analysis. Precision also depends on the amount and structure of usable data. In this trial, the primary outcome arm summaries list 10.0, 9.0, and 8.0 participants, respectively. Those outcome counts are not interchangeable with the overall enrollment figure, and the record does not explain all differences between them.

What is not documented?

The registry does not report the full statistical analysis plan, the target contrast for the mixed-model result, model diagnostics, missing-data methods, a confidence interval, or a multiplicity adjustment. It also does not state whether interim analyses were planned or performed, or provide a sample-size justification. These omissions prevent a complete independent appraisal of the inferential procedure. They should not be filled in by assuming standard choices: different reasonable model and multiplicity choices can lead to different interpretations.

For planning generally, power calculations describe the chance of detecting an effect under specified assumptions. A sample-size illustration based on an observed effect is not observed power and does not establish that the original trial was adequately designed for its question. The registry does not report the trial's own design justification or the assumptions behind its planning. Any such illustration should be read only as a planning scenario, not as evidence about the probability that the study did or did not detect an effect.

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("Control (Group A)", "8-h Afternoon-Evening Sleep (Group B)", "8-h Free Sleep (Group C)")
n   <- c(10, 9, 8)
m   <- c(338, 387, 428)
sp  <- c(68.7, 53.2, 66.7)
sd  <- sp      # posted standard deviations
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  Control (Group A)  10  338.000  68.700  288.855  387.145
CARM  8-h Afternoon-Evening Sleep (Group B)  9  387.000  53.200  346.107  427.893
CARM  8-h Free Sleep (Group C)  8  428.000  66.700  372.237  483.763

11. Analyses Reported by the Sponsor

The sponsor's reported analysis is a mixed models analysis across the three listed groups. It reports a mean difference (final values) of 90 and p < 0.05. The registry does not identify the specific contrast represented by that mean difference, provide a confidence interval, or describe the model in enough detail to reproduce the result.

A two-arm recalculation from the posted summary statistics is not available: the registry record characterizes the values as missing or not estimable for a two-arm comparison of this continuous endpoint. The descriptive arm means and standard deviations can be read directly, but they do not provide the missing model contrast or its uncertainty. There is no independent recalculated estimate to compare with the sponsor's reported model result.

The sponsor's result and the arm summaries serve different purposes. The summaries describe observed TST in each group; the model result is the registry's reported inferential finding. Since no discrepancy is identified, there is no conflicting recalculation to resolve. The reported p-value should still be interpreted with the stated limitations: its exact contrast is unclear, the confidence interval is not reported, and the analysis details are incomplete. The registry does not report whether the result was adjusted for multiple comparisons.

12. Multiplicity, Missing Data and Analysis Populations

The registry lists 15 outcome measures and reports 4 secondary outcomes. The primary outcome is TST. The secondary outcomes named in the record are fatigue and inter-shift recovery, subjective stress, subjective quality of life measured with WHOQOLBref, and salivary cortisol. The record notes that the cortisol aim was dropped when the study went fully remote due to COVID. It also reports 2 analyses, but does not explain their relationship to the listed outcomes or specify which analyses were primary, secondary, or exploratory.

When several outcomes, time points, or group contrasts are tested, the chance of at least one apparently positive result can increase. As a general statistical explanation, multiplicity procedures can control or describe that issue, but the appropriate method depends on the analysis plan and inferential goals. A multiplicity adjustment or testing hierarchy is not reported in the registry record. It is therefore unclear whether the reported p-value was adjusted for other tests, and it should not be treated as an unqualified assessment across all outcomes.

The primary outcome population is described as all participants with usable actigraphy data. The record does not report a detailed missing-data accounting, the number without usable recordings, reasons for missing or unusable data, or whether missingness differed among groups. It does not describe imputation, sensitivity analyses, or assumptions about the relationship between missingness and sleep duration. The posted means consequently describe the usable actigraphy data, not necessarily every participant enrolled or randomized.

The record also does not specify whether the reported analysis followed an intention-to-treat approach, how protocol deviations were handled, or whether the analysis population differed from the randomized population in other ways. The outcome arm counts are 10.0 for control, 9.0 for 8-h afternoon-evening sleep, and 8.0 for 8-h free sleep; the registry does not explain how those counts relate to enrollment and allocation. These limitations affect interpretation of both the descriptive summaries and the model result.

Information about planned interim analyses and the full statistical plan is not reported in the registry record. Their absence should be treated as uncertainty about the analysis documentation, not as evidence that a particular procedure was or was not used.

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.

Most frequent non-serious adverse event terms

The adverse-event record lists no serious adverse events and no deaths in any of the three groups. For other adverse events, it records 1 affected participant in control, 0 in 8-h afternoon-evening sleep, and 1 in 8-h free sleep. The corresponding at-risk counts are 10, 10, and 9. The registry lists no serious adverse-event terms.

These figures are counts, not estimates of comparative risk. The record does not provide event descriptions, timing, severity details for the other events, or an assessment of whether events were related to an intervention. The at-risk counts in the adverse-event record also differ from some primary-outcome counts; they refer to the safety reporting and should not be substituted for the actigraphy analysis population.

Zero recorded serious events and deaths in this record do not establish that an intervention is safe or that uncommon harms could be excluded. The study is small, and the registry does not provide enough detail to assess event ascertainment or follow-up. These safety figures should be read narrowly as the adverse-event counts reported, without inferring a causal relationship or making a treatment recommendation.

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

The posted data describe TST among participants with usable actigraphy and report a mixed-model result for the three-group study. The group means are 338.0 minutes for control, 387.0 minutes for 8-h afternoon-evening sleep, and 428.0 minutes for 8-h free sleep. The registry's model analysis reports a mean difference (final values) of 90 with p < 0.05. Because the specific contrast and confidence interval are not reported, the result cannot be translated into a precise pairwise effect with a clear uncertainty range.

These findings concern a measured sleep-duration endpoint in a field trial. They do not by themselves establish whether participants adopted the assigned timing, whether the strategy was feasible or acceptable, or whether fatigue, stress, quality of life, sleepiness, or performance changed. The registry's broader study aims and listed secondary outcomes are not accompanied by corresponding posted results in the record.

Nor do the results establish that one group experienced a clinically important benefit over another. Statistical evidence, the magnitude and precision of an effect, practical importance, and causal interpretation are related but distinct considerations. The incomplete model description and absence of a confidence interval limit assessment of magnitude and precision. Lack of masking and the behavioral nature of the interventions are design features; the record does not describe procedures to reduce the influence of expectations or behavioral differences in outcome assessment.

The results also cannot be generalized automatically beyond the study setting or population. The record describes older night workers in an operational environment, but does not report detailed participant characteristics or the reasons for study termination. The population with usable actigraphy may not represent everyone enrolled or all shift workers.

Overall, the registry records higher descriptive mean TST in the two sleep-instruction groups than in control, and reports a statistically significant mixed-model result. The reporting does not establish which pairwise comparison the model result represents, nor how precise it is. These are findings about the posted sleep-time data, not a complete evaluation of the interventions or their effects on health, safety, or work performance.

15. Why This Trial Matters Statistically

This trial is statistically interesting because it compares two distinct ways of structuring an 8-hour sleep opportunity with a control condition that received no sleep timing or duration instructions. One intervention specifies an afternoon-evening schedule, while the other leaves timing open. That design distinction makes the study question more nuanced than a simple instruction-versus-no-instruction comparison: timing and duration are related, but not identical, components of a behavioral strategy.

The outcome also links a practical intervention to an objectively recorded measure. Actigraphy-based TST can describe sleep duration between shifts, while the survey and focus-group components were designed to explore whether workers could and would adopt the intervention. Those questions matter statistically because an observed outcome among participants with usable recordings does not alone describe feasibility, adherence, or acceptability.

The reporting illustrates why an overall model result and group summaries need careful separation. The arm means describe the observed data, whereas the mixed-model result is inferential and its exact contrast is not identified. Without that contrast and a confidence interval, readers cannot assess the result as a specific comparison with a clear precision range. Reporting those details would make the statistical evidence easier to interpret and independently assess.

The record also highlights the role of analysis-population and missing-data descriptions. Usable actigraphy defined inclusion in the primary outcome, but the reasons and extent of unavailable data are not stated. When observed-data inclusion differs from the randomized population, transparent accounting helps readers understand the population represented by the estimates.

Finally, the study's broader aims are not fully answered by the posted primary outcome. Sleep duration is relevant to the intervention rationale, but the record does not report results for adoption, feasibility, acceptability, or the listed secondary outcomes. Distinguishing those questions prevents a narrow statistical finding from being taken as a complete account of a complex field trial.

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.

The registry provides limited detail about the analysis underlying the primary inferential result. It reports a mixed models analysis, a mean difference (final values) of 90, and p < 0.05, but does not identify the contrast, model terms, repeated-measure structure, covariate adjustment, or confidence interval. This makes the result difficult to reproduce or interpret as a specific comparison.

The primary outcome is restricted to participants with usable actigraphy data. The record does not explain the amount or causes of unavailable data or report analyses testing the effect of missingness. It also does not give a detailed account of how many nights contributed to each participant's average, despite describing an average for up to 3 intervention nights. This limits assessment of completeness and comparability of the endpoint.

The registry describes a randomized, unmasked, parallel-group design, but does not report sequence generation, allocation concealment, or the handling of protocol deviations. Because the interventions are behavioral and participants were told their group, masking participants was not part of the design as listed. The record does not describe whether outcome scoring or other assessments were masked.

The enrollment was 33, while the primary outcome summaries list arm counts of 10.0, 9.0, and 8.0. The relationship between enrollment, allocation, and usable actigraphy is not explained in the registry record. A sample-size justification, target detectable effect, and original design assumptions are also not reported. The registry does not state whether interim analyses occurred or describe a multiplicity plan.

The study status is terminated, but the reason is not reported in the registry record. Results for the survey, focus groups, and most listed secondary outcomes are not included in the posted results. The salivary cortisol aim is explicitly noted as dropped after the study went fully remote due to COVID. These factors limit conclusions about the broader feasibility and acceptability aims as well as the primary outcome's context.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: Sleep and Circadian Treatments for Shift Workers — NCT03813654.
• Registry record last updated 2025-08-15; status last verified 2025-07.
• Results first posted 2025-08-15.
• 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

The registry record describes a randomized, unmasked, parallel-group study of sleep and circadian treatments for shift workers, sponsored by Brigham and Women's Hospital. It lists an enrollment of 33 and a status of terminated. The study included a field trial, survey, and focus groups, with the field trial focused on sleep interventions for older night workers.

The primary endpoint was mean total sleep time in minutes, derived from actigraphy and averaged for up to 3 intervention nights. The posted group means were 338.0 for control, 387.0 for 8-h afternoon-evening sleep, and 428.0 for 8-h free sleep. The registry reports a mixed-model mean difference (final values) of 90 and p < 0.05, but does not identify the contrast or report a confidence interval.

Primary outcome inclusion was based on usable actigraphy data. A two-arm recalculation is not estimable from the posted summaries. The record does not report the full analysis plan, missing-data methods, or the reason for termination. It reports no serious adverse events or deaths and records 1 other adverse event in control and 1 in the free-sleep group.