Personalizing Sleep Interventions to Prevent Type 2 Diabetes in Community Dwelling Adults With Pre-Diabetes: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Personalizing Sleep Interventions to Prevent Type 2 Diabetes in Community Dwelling Adults With Pre-Diabetes (NCT03398902), a randomized na study. The primary outcome is Change in the Percent Time Glucose is ≥ 140 mg/dL From Pre- to Post-intervention.

2. Clinical Question

3. Trial Design

NCT03398902 is a completed, randomized, parallel-group study sponsored by New York University. The registry classifies its allocation as randomized and masking as none; no phase is listed beyond the registry entry of NA. The study compares a behavioral sleep-extension intervention with a habitual-sleep control condition. The record identifies pre-diabetes as the condition under study.

The intervention descriptions indicate that participants in both groups kept daily sleep diaries and took part in weekly telephone or videoconference contact. The sleep-extension group reviewed diaries with an instructor trained in Cognitive Behavioral Therapy for Insomnia. The habitual-sleep group was asked to keep usual bedtimes and wake times, with diaries reviewed by a study team member. The brief summary describes eight sessions and an eight-week intervention.

The registry lists study start as 2021-03-15, primary completion as 2023-09-18, and status as completed. It records enrollment of 393, while the brief summary says the randomized trial will include 150 adults. The posted outcome summaries concern 99 participants in the Sleep Intervention group and 92 in the Control group. The record does not clarify how the enrollment figure relates to the brief-summary target or to the participants represented in the outcome summaries.

The primary outcome description refers to assessment before and after intervention, with the timeframe given as baseline and post-treatment at approximately 12 weeks. The brief summary instead describes post-intervention collection at completion of the eight-week intervention. The registry does not clarify this difference in timing. Details such as the randomization sequence, allocation concealment, eligibility criteria, adherence definitions, and analysis population are not reported in 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.

Before assignment: Of 393 enrolled participants,191 met inclusion criteria and were randomized to treatment.

Overall Study

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 in the Percent Time Glucose is ≥ 140 mg/dL From Pre- to Post-intervention, measured over pre-treatment at baseline and post-treatment at ~12 weeks.

The primary endpoint is change in the percentage of time glucose was at least 140 mg/dL between pre-intervention and post-intervention assessment. The registry describes estimating this measure from approximately seven days of continuous glucose monitoring. It is a continuous outcome, summarized in the posted results as a mean with a standard deviation for each group.

Continuous glucose monitoring records glucose over time, allowing the endpoint to describe the portion of monitored time spent at or above the stated threshold rather than relying on a single measurement. Comparing change between groups is intended to distinguish the pattern over the study period in the sleep-extension group from the pattern in the habitual-sleep group. The registry does not explain the clinical rationale for selecting the 140 mg/dL threshold, define the precise calculation of percentage time, or state the direction used when subtracting baseline and post-intervention values. Those details are not reported in the registry record.

The endpoint description refers to a pre-to-post change, while its timeframe is listed as pre-treatment at baseline and post-treatment at approximately 12 weeks. The study brief describes completion of an eight-week intervention. The record does not reconcile these timing descriptions. The outcome should therefore be read according to the registry’s stated measure without assuming a more specific schedule or change convention than it records.

7. Results Posted in the Registry

• Sleep Intervention: mean 0.38 versus -0.49 on control; difference 0.87 (95% CI -0.27 to 2.01), standardized effect d = 0.21; z-test p = 0.135.

8. Statistical Interpretation

9. Statistical Methods Explained

The reported outcome is continuous: each participant’s value represents a change in a glucose-monitoring measure, and the group summaries use means and standard deviations. A mean describes the average recorded value within a group. A standard deviation describes how spread out individual values are around that group mean; it is not the uncertainty interval for the mean itself. Here, the posted means are 0.38 for Sleep Intervention and -0.49 for Control, with standard deviations of 4.74 and 3.22.

What does the mean difference describe?

The comparison is expressed as Sleep Intervention minus Control. Its reported mean difference is 0.87. This contrast summarizes the separation between the two group means on the measured scale. Because the endpoint is change, interpreting the sign as a favorable or unfavorable direction requires knowing whether change was calculated as post-intervention minus baseline or in the reverse order. That convention is not reported in the registry record. The difference is also unadjusted: it does not account for other variables in a statistical model.

How should the interval and p-value be read?

A confidence interval is a range of values compatible with the data and the assumptions used to calculate it. It communicates uncertainty around an estimate rather than giving a guarantee that the true effect lies within the particular interval. The reported interval runs from -0.272 to 2.012, crossing zero. That means the range includes no between-group difference as well as differences in either direction. The confidence level and exact interval method are not reported in the registry record.

A p-value describes how unusual the observed data, or data more extreme, would be under a specified null model and the test’s assumptions. The recalculated unadjusted p-value is 0.135. It is not the probability that the null hypothesis is true, and by itself it does not quantify the size or importance of the contrast. The registry does not report the sponsor’s primary inferential analysis, its model, or its prespecified decision threshold, so the recalculation should not be treated as a substitute for those details.

Why does variability matter?

More variability among individual measurements generally makes a group mean less precise when the group size is unchanged. Standard deviations of 4.74 and 3.22 indicate that the individual changes vary around their respective means. A pooled standard deviation can be used to express a standardized contrast, but such a measure depends on assumptions about how variability should be combined and does not restore the original glucose-monitoring scale. The registry’s recalculated standardized effect is not needed to interpret the raw mean difference, and it should not be mistaken for an adjusted estimate.

Continuous glucose monitoring can generate repeated measurements within a participant. The registry says the endpoint was estimated from approximately seven days of monitoring, but does not describe how the readings were processed into a participant-level percentage, how incomplete monitoring periods were treated, or how participant-level values were analyzed. General statistical practice distinguishes the repeated measurements used to construct an endpoint from the independent participants used in a between-group comparison. The specific procedures used in this trial are not reported.

What does randomization contribute?

Random allocation is intended to make treatment groups comparable on average with respect to both measured and unmeasured characteristics. It does not guarantee balance in a particular trial, and it does not remove uncertainty in the estimated difference. The registry labels allocation as randomized, but does not report the sequence-generation method, concealment procedures, stratification factors, or whether the posted analysis followed the randomized assignment for every participant.

The study is listed as parallel-group and unmasked. In a parallel design, participants are assigned to one group rather than receiving both conditions in sequence. With no masking listed, participants and study personnel may have known the assigned condition; the record does not describe who was aware of assignment or how outcome processing was protected from that knowledge. Because the primary endpoint is derived from glucose monitoring, the relevance of masking depends on measurement and analysis procedures that are not described.

These are general explanations of the statistical concepts. The exact analysis plan, assumptions checks, model specification, missing-data strategy, and handling of deviations are not reported in the registry record. Consequently, the calculations based on arm-level summaries are useful for describing the posted numbers, but they cannot answer questions that require participant-level data or details of the trial’s formal analysis.

10. Confidence Intervals and Precision

The reported mean difference is 0.87, comparing Sleep Intervention with Control. Its interval extends from -0.272 to 2.012. Because the lower endpoint is below zero and the upper endpoint is above zero, the interval includes no difference and values on either side of no difference. The interval is broad relative to the point estimate, so the estimate does not locate the between-group contrast precisely.

A confidence interval communicates uncertainty around an estimate under a specified method and its assumptions. It should not be read as a range containing every plausible clinical effect, nor as evidence that all values within it are equally likely. The confidence level and the method used for this interval are not reported in the registry record. The interval presented with the recalculated comparison is based on posted group summaries; it is not an adjusted model result.

The recalculated unadjusted p-value is 0.135. This gives another summary of the data relative to a null comparison, but it does not indicate the probability that the observed difference is due to chance, and it does not replace the interval’s description of uncertainty. The record does not report the primary analysis threshold or a sponsor-run inferential result.

Precision also depends on the quality and completeness of the underlying measurements. The registry describes approximately seven days of continuous glucose monitoring but does not report monitoring completeness, missing-data handling, or participant-level distributions. Those omissions limit assessment of how measurement and analysis choices may affect the interval. The posted estimate is therefore best read as an uncertain, unadjusted contrast on the registry’s recorded outcome scale, with its clinical direction and importance unresolved.

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 planning figures are participants per arm, not total event counts. They are planning illustrations based on the observed mean difference, rather than the trial’s original design calculations. The stated 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 to the stated fraction of the observed effect. The original design assumptions are not reported in the registry record.

For planning based on 100% of the observed effect, the figures are 346 participants per arm for 80% power and 463 participants per arm for 90% power. For 75% of the observed effect, they are 614 participants per arm for 80% power and 822 participants per arm for 90% power. For 50% of the observed effect, they are 1,381 participants per arm for 80% power and 1,849 participants per arm for 90% power.

In plain terms, power is the probability that a study would meet its specified statistical criterion if the assumed effect were the true effect and the planning assumptions held. The 80% and 90% columns describe two different target probabilities for detecting the assumed contrast; the higher target requires more participants in these illustrations. The smaller the effect selected for planning, the more participants are needed to distinguish it from random variation under the stated assumptions. The curve represents this same planning relationship rather than observed trial outcomes.

These figures are not observed power. They are hypothetical sample-size calculations using the posted effect estimate as a planning reference. They should not be read as evidence that the trial was designed around these assumptions or that its original enrollment target had a particular justification. The registry’s enrollment entry and brief-summary target are distinct entries, and the record does not explain their relationship to these illustrations.

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 = 0.870; pooled standard deviation σ = 4.079.
2. n = 2 × (1.96 + 0.842)² × 4.079² / 0.870² = 345.1, rounded up to 346 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 <- 0.87                      # observed difference in means
sigma <- 4.07946801179507                     # 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(3398902)
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 = 346 (80% power), 463 (90% power)
 75% of observed effect: n per arm = 614 (80% power), 822 (90% power)
 50% of observed effect: n per arm = 1,381 (80% power), 1,849 (90% power)
power.t.test: 346.1 
Simulated power at n = 346 per arm: 0.799 (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. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome measure and no posted secondary outcome measures. It also reports no analyses. A single listed primary outcome does not establish that the study tested only one hypothesis or that no additional outcomes were collected; it describes what is posted in the registry record. The record does not report a multiplicity strategy, such as an adjustment for multiple outcomes, time points, subgroup analyses, or repeated testing. It also does not give a prespecified hierarchy for interpreting results.

Multiplicity matters because conducting many statistical tests can increase the chance of finding an apparently unusual result even when no underlying effect exists. Whether adjustment is appropriate depends on the prespecified questions and analysis plan. Since those details are not reported, the posted primary outcome should not be assumed to be the only analysis considered by investigators, and no claim about multiplicity control can be made.

Missing-data handling is not reported in the registry record. The outcome relies on approximately seven days of continuous glucose monitoring, but the record does not say how many days of usable data were required, how incomplete monitoring was treated, whether missing outcome values were imputed, or whether analyses included only participants with complete measurements. The registry also does not state whether participants who discontinued or deviated from their assigned instructions remained in the analysis.

The analysis population is not reported. Although the study is labeled randomized, the record does not specify whether results were analyzed according to assigned group, according to intervention received, or under another rule. It does not describe exclusions after randomization or provide participant flow through outcome assessment. These distinctions can affect the interpretation of a group comparison, particularly when adherence or outcome availability differs between groups.

These omissions do not show that the study failed to plan or perform these procedures; they mean that the details are not reported in the registry record. The recalculated contrast from group summaries cannot reconstruct the missing analysis plan, account for missing observations, or address multiplicity.

14. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent non-serious adverse event terms

The registry lists zero serious adverse events among 99 participants at risk in the Sleep Intervention group and zero among 92 participants at risk in Control. It also lists zero deaths in both groups. These are the recorded counts for the groups shown; they do not establish that no events occurred outside the registry’s reporting categories or observation period.

The record lists 13 other adverse events in Sleep Intervention and 0 in Control. It does not provide terms for those events in the serious-event list, which is empty, and does not describe the definition of the other-event category, event timing, severity, or whether events were considered related to the intervention. These counts should not be interpreted as evidence of causation or as a complete characterization of safety.

Adverse-event counts are descriptive. A zero count in a group means no event in that category was recorded for that group in the registry entry; it does not prove that the underlying risk is zero. The record does not report a safety analysis plan or further details about collection and follow-up, so the figures are limited to the posted counts.

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

The posted result is a comparison of average pre-to-post changes in the percentage of monitored time glucose was at least 140 mg/dL. The mean was 0.38 in Sleep Intervention and -0.49 in Control, and the unadjusted recalculated difference was 0.87 for Sleep Intervention minus Control. The interval crosses zero, while the recalculated p-value is 0.135. Together, these statistics describe an uncertain contrast rather than a precise estimate of a single direction or magnitude.

The result does not, by itself, establish whether the intervention improved or worsened the endpoint. The registry does not specify the sign convention for the change score, and it does not identify a clinically important difference. The endpoint also captures a glucose-monitoring measure over an approximately seven-day period; the posted summaries do not establish how that measure relates to longer-term outcomes.

The statistics do not establish that the groups differed because of the sleep instructions, nor do they rule out a meaningful effect in either direction. The record identifies randomized allocation, but does not report key design and analysis details such as allocation procedures, baseline characteristics, adherence, missing-data handling, and the analysis population. The unadjusted comparison is based on group-level summaries and cannot account for those unreported factors.

Likewise, the result does not resolve the broader question of how a sleep intervention might affect glucose regulation in all adults with pre-diabetes. The registry describes a specific community-dwelling population and a particular diary-supported sleep-extension program, but provides no subgroup results or details that would establish how broadly the findings apply. The mismatch between the brief’s eight-week intervention and the endpoint timeframe of approximately 12 weeks is also not clarified.

The appropriate reading is narrow: the registry posts group means and variability, along with a recalculated unadjusted contrast and interval. The point estimate is positive under the recorded comparison, but the interval includes zero and the direction of the underlying change scale is not defined. These data alone do not warrant a medical recommendation or a definitive claim about treatment effect.

16. Why This Trial Matters Statistically

This trial is statistically interesting because it connects a behavioral intervention with an objectively monitored continuous outcome. Continuous glucose monitoring provides repeated glucose readings from which the registry says the percentage of time at or above 140 mg/dL was estimated. The analysis question is not simply whether one group had a different value at one moment; it concerns a change over time compared between randomized groups.

The contrast also illustrates why the estimate, its uncertainty, and the measurement definition must be considered together. A positive mean difference is not self-interpreting when the registry does not state the direction used to calculate change. The interval crosses zero, showing that the posted summary estimate is not precise enough to exclude differences on either side of zero under its calculation. Statistical interpretation depends on both the numerical estimate and clarity about what that estimate represents.

Differences between the registry’s enrollment entry, brief-summary target, posted group sizes, and timing descriptions are relevant to transparent reporting. These entries may describe different stages or aspects of the study, but the record does not explain how they relate. A clear account of participant flow, outcome timing, analysis populations, and missing monitoring data would help readers assess how the posted comparison arose.

The planning illustration also shows the dependence of sample-size requirements on the effect assumed for detection and on the desired power. These are hypothetical calculations based on the observed contrast, not evidence about the original design. Keeping that distinction clear prevents a planning exercise from being misread as an evaluation of the trial’s achieved statistical power.

Overall, the record offers a useful example of how randomized design, continuous outcome summaries, uncertainty intervals, and incomplete analysis reporting shape interpretation. The central statistical lesson is to separate what the posted estimate says numerically from what is not established about its direction, precision, or clinical meaning.

17. 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 trial-specific reporting gaps limit interpretation. The registry’s brief summary states that the randomized trial will include 150 adults, while the enrollment entry is 393 and the posted outcome groups include 99 and 92 participants. The record does not explain whether these figures refer to different enrollment stages, analysis populations, or other distinctions. Without participant flow, the relationship among them cannot be resolved.

The timing descriptions also differ: the brief describes an eight-week intervention and assessment on completion, whereas the primary outcome timeframe says post-treatment at approximately 12 weeks. The record does not explain whether this reflects a follow-up period, a difference in wording, or another schedule. The meaning of the change-score sign is also unspecified, limiting directional interpretation.

Details needed to assess the intervention and analysis are not reported in the registry record. These include the sleep-extension targets, adherence, randomization procedures, baseline characteristics, analysis population, missing-data approach, outcome-processing rules, and the sponsor’s statistical analysis. The registry lists no posted analyses, and the unadjusted comparison is recalculated from arm-level summaries. The confidence level for the reported interval is not stated.

Finally, the record gives limited safety detail beyond event counts and does not define the other-adverse-event category. It also does not provide secondary outcome results. These reporting limits do not show that the procedures or information were absent from the full study; they restrict what can be concluded from the registry record.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Personalizing Sleep Interventions to Prevent Type 2 Diabetes in Community Dwelling Adults With Pre-Diabetes — NCT03398902.
• Registry record last updated 2025-09-04; status last verified 2023-11.
• Results first posted 2025-09-04.
• 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.

21. Record Summary

NCT03398902 is a completed randomized, parallel-group study sponsored by New York University. It compares a sleep-extension intervention with instructions to maintain habitual sleep among community-dwelling adults with pre-diabetes. The registry lists no masking and identifies the primary outcome as change in percentage of time glucose was at least 140 mg/dL, estimated using approximately seven days of continuous glucose monitoring.

The posted group means are 0.38 for Sleep Intervention and -0.49 for Control. The unadjusted recalculated mean difference is 0.87, with an interval from -0.272 to 2.012 and p-value 0.135. The interval crosses zero, and the registry does not specify the change-score direction or confidence level.

The record lists zero serious adverse events and zero deaths in both groups. It does not report the primary analysis methods, missing-data procedures, or analysis population. It also contains different enrollment and timing descriptions that it does not reconcile.